platform/kernel/linux-rpi.git
2 years agouapi/linux/stddef.h: Add include guards
Tadeusz Struk [Tue, 29 Mar 2022 17:12:52 +0000 (10:12 -0700)]
uapi/linux/stddef.h: Add include guards

[ Upstream commit 55037ed7bdc62151a726f5685f88afa6a82959b1 ]

Add include guard wrapper define to uapi/linux/stddef.h to prevent macro
redefinition errors when stddef.h is included more than once. This was not
needed before since the only contents already used a redefinition test.

Signed-off-by: Tadeusz Struk <tadeusz.struk@linaro.org>
Link: https://lore.kernel.org/r/20220329171252.57279-1-tadeusz.struk@linaro.org
Fixes: 50d7bd38c3aa ("stddef: Introduce struct_group() helper macro")
Cc: stable@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agostddef: Introduce DECLARE_FLEX_ARRAY() helper
Kees Cook [Mon, 9 Aug 2021 18:21:23 +0000 (11:21 -0700)]
stddef: Introduce DECLARE_FLEX_ARRAY() helper

[ Upstream commit 3080ea5553cc909b000d1f1d964a9041962f2c5b ]

There are many places where kernel code wants to have several different
typed trailing flexible arrays. This would normally be done with multiple
flexible arrays in a union, but since GCC and Clang don't (on the surface)
allow this, there have been many open-coded workarounds, usually involving
neighboring 0-element arrays at the end of a structure. For example,
instead of something like this:

struct thing {
...
union {
struct type1 foo[];
struct type2 bar[];
};
};

code works around the compiler with:

struct thing {
...
struct type1 foo[0];
struct type2 bar[];
};

Another case is when a flexible array is wanted as the single member
within a struct (which itself is usually in a union). For example, this
would be worked around as:

union many {
...
struct {
struct type3 baz[0];
};
};

These kinds of work-arounds cause problems with size checks against such
zero-element arrays (for example when building with -Warray-bounds and
-Wzero-length-bounds, and with the coming FORTIFY_SOURCE improvements),
so they must all be converted to "real" flexible arrays, avoiding warnings
like this:

fs/hpfs/anode.c: In function 'hpfs_add_sector_to_btree':
fs/hpfs/anode.c:209:27: warning: array subscript 0 is outside the bounds of an interior zero-length array 'struct bplus_internal_node[0]' [-Wzero-length-bounds]
  209 |    anode->btree.u.internal[0].down = cpu_to_le32(a);
      |    ~~~~~~~~~~~~~~~~~~~~~~~^~~
In file included from fs/hpfs/hpfs_fn.h:26,
                 from fs/hpfs/anode.c:10:
fs/hpfs/hpfs.h:412:32: note: while referencing 'internal'
  412 |     struct bplus_internal_node internal[0]; /* (internal) 2-word entries giving
      |                                ^~~~~~~~

drivers/net/can/usb/etas_es58x/es58x_fd.c: In function 'es58x_fd_tx_can_msg':
drivers/net/can/usb/etas_es58x/es58x_fd.c:360:35: warning: array subscript 65535 is outside the bounds of an interior zero-length array 'u8[0]' {aka 'unsigned char[]'} [-Wzero-length-bounds]
  360 |  tx_can_msg = (typeof(tx_can_msg))&es58x_fd_urb_cmd->raw_msg[msg_len];
      |                                   ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from drivers/net/can/usb/etas_es58x/es58x_core.h:22,
                 from drivers/net/can/usb/etas_es58x/es58x_fd.c:17:
drivers/net/can/usb/etas_es58x/es58x_fd.h:231:6: note: while referencing 'raw_msg'
  231 |   u8 raw_msg[0];
      |      ^~~~~~~

However, it _is_ entirely possible to have one or more flexible arrays
in a struct or union: it just has to be in another struct. And since it
cannot be alone in a struct, such a struct must have at least 1 other
named member -- but that member can be zero sized. Wrap all this nonsense
into the new DECLARE_FLEX_ARRAY() in support of having flexible arrays
in unions (or alone in a struct).

As with struct_group(), since this is needed in UAPI headers as well,
implement the core there, with a non-UAPI wrapper.

Additionally update kernel-doc to understand its existence.

https://github.com/KSPP/linux/issues/137

Cc: Arnd Bergmann <arnd@arndb.de>
Cc: "Gustavo A. R. Silva" <gustavoars@kernel.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobus: mhi: Fix pm_state conversion to string
Paul Davey [Tue, 1 Mar 2022 16:03:00 +0000 (21:33 +0530)]
bus: mhi: Fix pm_state conversion to string

[ Upstream commit 64f93a9a27c1970fa8ee5ffc5a6ae2bda477ec5b ]

On big endian architectures the mhi debugfs files which report pm state
give "Invalid State" for all states.  This is caused by using
find_last_bit which takes an unsigned long* while the state is passed in
as an enum mhi_pm_state which will be of int size.

Fix by using __fls to pass the value of state instead of find_last_bit.

Also the current API expects "mhi_pm_state" enumerator as the function
argument but the function only works with bitmasks. So as Alex suggested,
let's change the argument to u32 to avoid confusion.

Fixes: a6e2e3522f29 ("bus: mhi: core: Add support for PM state transitions")
Cc: stable@vger.kernel.org
[mani: changed the function argument to u32]
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
Reviewed-by: Hemant Kumar <hemantk@codeaurora.org>
Reviewed-by: Alex Elder <elder@linaro.org>
Signed-off-by: Paul Davey <paul.davey@alliedtelesis.co.nz>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20220301160308.107452-3-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobus: mhi: core: Use correctly sized arguments for bit field
Kees Cook [Thu, 16 Dec 2021 08:12:26 +0000 (13:42 +0530)]
bus: mhi: core: Use correctly sized arguments for bit field

[ Upstream commit 5a717e93239fc373a314e03e45c43b62ebea1b26 ]

The find.h APIs are designed to be used only on unsigned long arguments.
This can technically result in a over-read, but it is harmless in this
case. Regardless, fix it to avoid the warning seen under -Warray-bounds,
which we'd like to enable globally:

In file included from ./include/linux/bitmap.h:9,
                 from ./include/linux/cpumask.h:12,
                 from ./arch/x86/include/asm/cpumask.h:5,
                 from ./arch/x86/include/asm/msr.h:11,
                 from ./arch/x86/include/asm/processor.h:22,
                 from ./arch/x86/include/asm/cpufeature.h:5,
                 from ./arch/x86/include/asm/thread_info.h:53,
                 from ./include/linux/thread_info.h:60,
                 from ./arch/x86/include/asm/preempt.h:7,
                 from ./include/linux/preempt.h:78,
                 from ./include/linux/spinlock.h:55,
                 from ./include/linux/wait.h:9,
                 from ./include/linux/wait_bit.h:8,
                 from ./include/linux/fs.h:6,
                 from ./include/linux/debugfs.h:15,
                 from drivers/bus/mhi/core/init.c:7:
drivers/bus/mhi/core/init.c: In function 'to_mhi_pm_state_str':
./include/linux/find.h:187:37: warning: array subscript 'long unsigned int[0]' is partly outside array bounds of 'enum mhi_pm_state[1]' [-Warray-bounds]
  187 |                 unsigned long val = *addr & GENMASK(size - 1, 0);
      |                                     ^~~~~
drivers/bus/mhi/core/init.c:80:51: note: while referencing 'state'
   80 | const char *to_mhi_pm_state_str(enum mhi_pm_state state)
      |                                 ~~~~~~~~~~~~~~~~~~^~~~~

Link: https://lore.kernel.org/r/20211215232446.2069794-1-keescook@chromium.org
[mani: changed the variable name "bits" to "pm_state"]
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Link: https://lore.kernel.org/r/20211216081227.237749-10-manivannan.sadhasivam@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoserial: sc16is7xx: Clear RS485 bits in the shutdown
Hui Wang [Tue, 8 Mar 2022 11:00:42 +0000 (19:00 +0800)]
serial: sc16is7xx: Clear RS485 bits in the shutdown

[ Upstream commit 927728a34f11b5a27f4610bdb7068317d6fdc72a ]

We tested RS485 function on an EVB which has SC16IS752, after
finishing the test, we started the RS232 function test, but found the
RTS is still working in the RS485 mode.

That is because both startup and shutdown call port_update() to set
the EFCR_REG, this will not clear the RS485 bits once the bits are set
in the reconf_rs485(). To fix it, clear the RS485 bits in shutdown.

Cc: <stable@vger.kernel.org>
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Link: https://lore.kernel.org/r/20220308110042.108451-1-hui.wang@canonical.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agopowerpc/tm: Fix more userspace r13 corruption
Nicholas Piggin [Fri, 11 Mar 2022 02:47:33 +0000 (12:47 +1000)]
powerpc/tm: Fix more userspace r13 corruption

[ Upstream commit 9d71165d3934e607070c4e48458c0cf161b1baea ]

Commit cf13435b730a ("powerpc/tm: Fix userspace r13 corruption") fixes a
problem in treclaim where a SLB miss can occur on the
thread_struct->ckpt_regs while SCRATCH0 is live with the saved user r13
value, clobbering it with the kernel r13 and ultimately resulting in
kernel r13 being stored in ckpt_regs.

There is an equivalent problem in trechkpt where the user r13 value is
loaded into r13 from chkpt_regs to be recheckpointed, but a SLB miss
could occur on ckpt_regs accesses after that, which will result in r13
being clobbered with a kernel value and that will get recheckpointed and
then restored to user registers.

The same memory page is accessed right before this critical window where
a SLB miss could cause corruption, so hitting the bug requires the SLB
entry be removed within a small window of instructions, which is
possible if a SLB related MCE hits there. PAPR also permits the
hypervisor to discard this SLB entry (because slb_shadow->persistent is
only set to SLB_NUM_BOLTED) although it's not known whether any
implementations would do this (KVM does not). So this is an extremely
unlikely bug, only found by inspection.

Fix this by also storing user r13 in a temporary location on the kernel
stack and don't change the r13 register from kernel r13 until the RI=0
critical section that does not fault.

The SCRATCH0 change is not strictly part of the fix, it's only used in
the RI=0 section so it does not have the same problem as the previous
SCRATCH0 bug.

Fixes: 98ae22e15b43 ("powerpc: Add helper functions for transactional memory context switching")
Cc: stable@vger.kernel.org # v3.9+
Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Acked-by: Michael Neuling <mikey@neuling.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20220311024733.48926-1-npiggin@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agopowerpc: flexible GPR range save/restore macros
Nicholas Piggin [Fri, 22 Oct 2021 06:13:22 +0000 (16:13 +1000)]
powerpc: flexible GPR range save/restore macros

[ Upstream commit aebd1fb45c622e9a2b06fb70665d084d3a8d6c78 ]

Introduce macros that operate on a (start, end) range of GPRs, which
reduces lines of code and need to do mental arithmetic while reading the
code.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
Reviewed-by: Segher Boessenkool <segher@kernel.crashing.org>
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20211022061322.2671178-1-npiggin@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agopowerpc/32: Don't use lmw/stmw for saving/restoring non volatile regs
Christophe Leroy [Mon, 23 Aug 2021 15:29:12 +0000 (15:29 +0000)]
powerpc/32: Don't use lmw/stmw for saving/restoring non volatile regs

[ Upstream commit a85c728cb5e12216c19ae5878980c2cbbbf8616d ]

Instructions lmw/stmw are interesting for functions that are rarely
used and not in the cache, because only one instruction is to be
copied into the instruction cache instead of 19. However those
instruction are less performant than 19x raw lwz/stw as they require
synchronisation plus one additional cycle.

SAVE_NVGPRS / REST_NVGPRS are used in only a few places which are
mostly in interrupts entries/exits and in task switch so they are
likely already in the cache.

Using standard lwz improves null_syscall selftest by:
- 10 cycles on mpc832x.
- 2 cycles on mpc8xx.

Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/316c543b8906712c108985c8463eec09c8db577b.1629732542.git.christophe.leroy@csgroup.eu
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoscsi: qla2xxx: Fix loss of NVMe namespaces after driver reload test
Arun Easi [Thu, 10 Mar 2022 09:25:54 +0000 (01:25 -0800)]
scsi: qla2xxx: Fix loss of NVMe namespaces after driver reload test

[ Upstream commit db212f2eb3fb7f546366777e93c8f54614d39269 ]

Driver registration of localport can race when it happens at the remote
port discovery time. Fix this by calling the registration under a mutex.

Link: https://lore.kernel.org/r/20220310092604.22950-4-njavali@marvell.com
Fixes: e84067d74301 ("scsi: qla2xxx: Add FC-NVMe F/W initialization and transport registration")
Cc: stable@vger.kernel.org
Reported-by: Marco Patalano <mpatalan@redhat.com>
Tested-by: Marco Patalano <mpatalan@redhat.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoKVM: s390x: fix SCK locking
Claudio Imbrenda [Tue, 1 Mar 2022 14:33:40 +0000 (15:33 +0100)]
KVM: s390x: fix SCK locking

[ Upstream commit c0573ba5c5a2244dc02060b1f374d4593c1d20b7 ]

When handling the SCK instruction, the kvm lock is taken, even though
the vcpu lock is already being held. The normal locking order is kvm
lock first and then vcpu lock. This is can (and in some circumstances
does) lead to deadlocks.

The function kvm_s390_set_tod_clock is called both by the SCK handler
and by some IOCTLs to set the clock. The IOCTLs will not hold the vcpu
lock, so they can safely take the kvm lock. The SCK handler holds the
vcpu lock, but will also somehow need to acquire the kvm lock without
relinquishing the vcpu lock.

The solution is to factor out the code to set the clock, and provide
two wrappers. One is called like the original function and does the
locking, the other is called kvm_s390_try_set_tod_clock and uses
trylock to try to acquire the kvm lock. This new wrapper is then used
in the SCK handler. If locking fails, -EAGAIN is returned, which is
eventually propagated to userspace, thus also freeing the vcpu lock and
allowing for forward progress.

This is not the most efficient or elegant way to solve this issue, but
the SCK instruction is deprecated and its performance is not critical.

The goal of this patch is just to provide a simple but correct way to
fix the bug.

Fixes: 6a3f95a6b04c ("KVM: s390: Intercept SCK instruction")
Signed-off-by: Claudio Imbrenda <imbrenda@linux.ibm.com>
Reviewed-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Reviewed-by: Janis Schoetterl-Glausch <scgl@linux.ibm.com>
Link: https://lore.kernel.org/r/20220301143340.111129-1-imbrenda@linux.ibm.com
Cc: stable@vger.kernel.org
Signed-off-by: Christian Borntraeger <borntraeger@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: don't access possibly stale fs_info data in device_list_add
Dongliang Mu [Thu, 3 Mar 2022 14:40:27 +0000 (22:40 +0800)]
btrfs: don't access possibly stale fs_info data in device_list_add

[ Upstream commit 79c9234ba596e903907de20573fd4bcc85315b06 ]

Syzbot reported a possible use-after-free in printing information
in device_list_add.

Very similar with the bug fixed by commit 0697d9a61099 ("btrfs: don't
access possibly stale fs_info data for printing duplicate device"),
but this time the use occurs in btrfs_info_in_rcu.

  Call Trace:
   kasan_report.cold+0x83/0xdf mm/kasan/report.c:459
   btrfs_printk+0x395/0x425 fs/btrfs/super.c:244
   device_list_add.cold+0xd7/0x2ed fs/btrfs/volumes.c:957
   btrfs_scan_one_device+0x4c7/0x5c0 fs/btrfs/volumes.c:1387
   btrfs_control_ioctl+0x12a/0x2d0 fs/btrfs/super.c:2409
   vfs_ioctl fs/ioctl.c:51 [inline]
   __do_sys_ioctl fs/ioctl.c:874 [inline]
   __se_sys_ioctl fs/ioctl.c:860 [inline]
   __x64_sys_ioctl+0x193/0x200 fs/ioctl.c:860
   do_syscall_x64 arch/x86/entry/common.c:50 [inline]
   do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
   entry_SYSCALL_64_after_hwframe+0x44/0xae

Fix this by modifying device->fs_info to NULL too.

Reported-and-tested-by: syzbot+82650a4e0ed38f218363@syzkaller.appspotmail.com
CC: stable@vger.kernel.org # 4.19+
Signed-off-by: Dongliang Mu <mudongliangabcd@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoKVM: use __vcalloc for very large allocations
Paolo Bonzini [Tue, 8 Mar 2022 09:49:37 +0000 (04:49 -0500)]
KVM: use __vcalloc for very large allocations

[ Upstream commit 37b2a6510a48ca361ced679f92682b7b7d7d0330 ]

Allocations whose size is related to the memslot size can be arbitrarily
large.  Do not use kvzalloc/kvcalloc, as those are limited to "not crazy"
sizes that fit in 32 bits.

Cc: stable@vger.kernel.org
Fixes: 7661809d493b ("mm: don't allow oversized kvmalloc() calls")
Reviewed-by: David Hildenbrand <david@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomm: vmalloc: introduce array allocation functions
Paolo Bonzini [Tue, 8 Mar 2022 09:47:22 +0000 (04:47 -0500)]
mm: vmalloc: introduce array allocation functions

[ Upstream commit a8749a35c39903120ec421ef2525acc8e0daa55c ]

Linux has dozens of occurrences of vmalloc(array_size()) and
vzalloc(array_size()).  Allow to simplify the code by providing
vmalloc_array and vcalloc, as well as the underscored variants that let
the caller specify the GFP flags.

Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoCompiler Attributes: add __alloc_size() for better bounds checking
Kees Cook [Fri, 5 Nov 2021 20:36:19 +0000 (13:36 -0700)]
Compiler Attributes: add __alloc_size() for better bounds checking

[ Upstream commit 86cffecdeaa278444870c8745ab166a65865dbf0 ]

GCC and Clang can use the "alloc_size" attribute to better inform the
results of __builtin_object_size() (for compile-time constant values).
Clang can additionally use alloc_size to inform the results of
__builtin_dynamic_object_size() (for run-time values).

Because GCC sees the frequent use of struct_size() as an allocator size
argument, and notices it can return SIZE_MAX (the overflow indication),
it complains about these call sites overflowing (since SIZE_MAX is
greater than the default -Walloc-size-larger-than=PTRDIFF_MAX).  This
isn't helpful since we already know a SIZE_MAX will be caught at
run-time (this was an intentional design).  To deal with this, we must
disable this check as it is both a false positive and redundant.  (Clang
does not have this warning option.)

Unfortunately, just checking the -Wno-alloc-size-larger-than is not
sufficient to make the __alloc_size attribute behave correctly under
older GCC versions.  The attribute itself must be disabled in those
situations too, as there appears to be no way to reliably silence the
SIZE_MAX constant expression cases for GCC versions less than 9.1:

   In file included from ./include/linux/resource_ext.h:11,
                    from ./include/linux/pci.h:40,
                    from drivers/net/ethernet/intel/ixgbe/ixgbe.h:9,
                    from drivers/net/ethernet/intel/ixgbe/ixgbe_lib.c:4:
   In function 'kmalloc_node',
       inlined from 'ixgbe_alloc_q_vector' at ./include/linux/slab.h:743:9:
   ./include/linux/slab.h:618:9: error: argument 1 value '18446744073709551615' exceeds maximum object size 9223372036854775807 [-Werror=alloc-size-larger-than=]
     return __kmalloc_node(size, flags, node);
            ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
   ./include/linux/slab.h: In function 'ixgbe_alloc_q_vector':
   ./include/linux/slab.h:455:7: note: in a call to allocation function '__kmalloc_node' declared here
    void *__kmalloc_node(size_t size, gfp_t flags, int node) __assume_slab_alignment __malloc;
          ^~~~~~~~~~~~~~

Specifically:
 '-Wno-alloc-size-larger-than' is not correctly handled by GCC < 9.1
    https://godbolt.org/z/hqsfG7q84 (doesn't disable)
    https://godbolt.org/z/P9jdrPTYh (doesn't admit to not knowing about option)
    https://godbolt.org/z/465TPMWKb (only warns when other warnings appear)

 '-Walloc-size-larger-than=18446744073709551615' is not handled by GCC < 8.2
    https://godbolt.org/z/73hh1EPxz (ignores numeric value)

Since anything marked with __alloc_size would also qualify for marking
with __malloc, just include __malloc along with it to avoid redundant
markings.  (Suggested by Linus Torvalds.)

Finally, make sure checkpatch.pl doesn't get confused about finding the
__alloc_size attribute on functions.  (Thanks to Joe Perches.)

Link: https://lkml.kernel.org/r/20210930222704.2631604-3-keescook@chromium.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Tested-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Andy Whitcroft <apw@canonical.com>
Cc: Christoph Lameter <cl@linux.com>
Cc: Daniel Micay <danielmicay@gmail.com>
Cc: David Rientjes <rientjes@google.com>
Cc: Dennis Zhou <dennis@kernel.org>
Cc: Dwaipayan Ray <dwaipayanray1@gmail.com>
Cc: Joe Perches <joe@perches.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Cc: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: Tejun Heo <tj@kernel.org>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Alexandre Bounine <alex.bou9@gmail.com>
Cc: Gustavo A. R. Silva <gustavoars@kernel.org>
Cc: Ira Weiny <ira.weiny@intel.com>
Cc: Jing Xiangfeng <jingxiangfeng@huawei.com>
Cc: John Hubbard <jhubbard@nvidia.com>
Cc: kernel test robot <lkp@intel.com>
Cc: Matt Porter <mporter@kernel.crashing.org>
Cc: Miguel Ojeda <ojeda@kernel.org>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Souptick Joarder <jrdr.linux@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomtd: spi-nor: Skip erase logic when SPI_NOR_NO_ERASE is set
Tudor Ambarus [Mon, 28 Feb 2022 16:33:34 +0000 (18:33 +0200)]
mtd: spi-nor: Skip erase logic when SPI_NOR_NO_ERASE is set

[ Upstream commit 151c6b49d679872d6fc0b50e0ad96303091694a2 ]

Even if SPI_NOR_NO_ERASE was set, one could still send erase opcodes
to the flash. It is not recommended to send unsupported opcodes to
flashes. Fix the logic and do not set mtd->_erase when SPI_NOR_NO_ERASE
is specified. With this users will not be able to issue erase opcodes to
flashes and instead they will recive an -ENOTSUPP error.

Fixes: b199489d37b2 ("mtd: spi-nor: add the framework for SPI NOR")
Signed-off-by: Tudor Ambarus <tudor.ambarus@microchip.com>
Reviewed-by: Michael Walle <michael@walle.cc>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20220228163334.277730-1-tudor.ambarus@microchip.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobatman-adv: Use netif_rx().
Sebastian Andrzej Siewior [Sun, 6 Mar 2022 21:57:48 +0000 (22:57 +0100)]
batman-adv: Use netif_rx().

[ Upstream commit 94da81e2fc4285db373fe9a1eb012c2ee205b110 ]

Since commit
   baebdf48c3600 ("net: dev: Makes sure netif_rx() can be invoked in any context.")

the function netif_rx() can be used in preemptible/thread context as
well as in interrupt context.

Use netif_rx().

Cc: Antonio Quartulli <a@unstable.cc>
Cc: Marek Lindner <mareklindner@neomailbox.ch>
Cc: Simon Wunderlich <sw@simonwunderlich.de>
Cc: Sven Eckelmann <sven@narfation.org>
Cc: b.a.t.m.a.n@lists.open-mesh.org
Signed-off-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Acked-by: Sven Eckelmann <sven@narfation.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoiio: accel: mma8452: use the correct logic to get mma8452_data
Haibo Chen [Tue, 22 Feb 2022 02:42:21 +0000 (10:42 +0800)]
iio: accel: mma8452: use the correct logic to get mma8452_data

[ Upstream commit c87b7b12f48db86ac9909894f4dc0107d7df6375 ]

The original logic to get mma8452_data is wrong, the *dev point to
the device belong to iio_dev. we can't use this dev to find the
correct i2c_client. The original logic happen to work because it
finally use dev->driver_data to get iio_dev. Here use the API
to_i2c_client() is wrong and make reader confuse. To correct the
logic, it should be like this

  struct mma8452_data *data = iio_priv(dev_get_drvdata(dev));

But after commit 8b7651f25962 ("iio: iio_device_alloc(): Remove
unnecessary self drvdata"), the upper logic also can't work.
When try to show the avialable scale in userspace, will meet kernel
dump, kernel handle NULL pointer dereference.

So use dev_to_iio_dev() to correct the logic.

Dual fixes tags as the second reflects when the bug was exposed, whilst
the first reflects when the original bug was introduced.

Fixes: c3cdd6e48e35 ("iio: mma8452: refactor for seperating chip specific data")
Fixes: 8b7651f25962 ("iio: iio_device_alloc(): Remove unnecessary self drvdata")
Signed-off-by: Haibo Chen <haibo.chen@nxp.com>
Reviewed-by: Martin Kepplinger <martink@posteo.de>
Cc: <Stable@vger.kernel.org>
Link: https://lore.kernel.org/r/1645497741-5402-1-git-send-email-haibo.chen@nxp.com
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoriscv/mm: Add XIP_FIXUP for riscv_pfn_base
Palmer Dabbelt [Fri, 4 Feb 2022 21:14:08 +0000 (13:14 -0800)]
riscv/mm: Add XIP_FIXUP for riscv_pfn_base

[ Upstream commit ca0cb9a60f6d86d4b2139c6f393a78f39edcd7cb ]

This manifests as a crash early in boot on VexRiscv.

Signed-off-by: Myrtle Shah <gatecat@ds0.me>
[Palmer: split commit]
Fixes: 44c922572952 ("RISC-V: enable XIP")
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoNFSD: COMMIT operations must not return NFS?ERR_INVAL
Chuck Lever [Mon, 24 Jan 2022 20:50:31 +0000 (15:50 -0500)]
NFSD: COMMIT operations must not return NFS?ERR_INVAL

[ Upstream commit 3f965021c8bc38965ecb1924f570c4842b33d408 ]

Since, well, forever, the Linux NFS server's nfsd_commit() function
has returned nfserr_inval when the passed-in byte range arguments
were non-sensical.

However, according to RFC 1813 section 3.3.21, NFSv3 COMMIT requests
are permitted to return only the following non-zero status codes:

      NFS3ERR_IO
      NFS3ERR_STALE
      NFS3ERR_BADHANDLE
      NFS3ERR_SERVERFAULT

NFS3ERR_INVAL is not included in that list. Likewise, NFS4ERR_INVAL
is not listed in the COMMIT row of Table 6 in RFC 8881.

RFC 7530 does permit COMMIT to return NFS4ERR_INVAL, but does not
specify when it can or should be used.

Instead of dropping or failing a COMMIT request in a byte range that
is not supported, turn it into a valid request by treating one or
both arguments as zero. Offset zero means start-of-file, count zero
means until-end-of-file, so we only ever extend the commit range.
NFS servers are always allowed to commit more and sooner than
requested.

The range check is no longer bounded by NFS_OFFSET_MAX, but rather
by the value that is returned in the maxfilesize field of the NFSv3
FSINFO procedure or the NFSv4 maxfilesize file attribute.

Note that this change results in a new pynfs failure:

CMT4     st_commit.testCommitOverflow                             : RUNNING
CMT4     st_commit.testCommitOverflow                             : FAILURE
           COMMIT with offset + count overflow should return
           NFS4ERR_INVAL, instead got NFS4_OK

IMO the test is not correct as written: RFC 8881 does not allow the
COMMIT operation to return NFS4ERR_INVAL.

Reported-by: Dan Aloni <dan.aloni@vastdata.com>
Cc: stable@vger.kernel.org
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Bruce Fields <bfields@fieldses.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoNFSD: De-duplicate net_generic(nf->nf_net, nfsd_net_id)
Chuck Lever [Tue, 28 Dec 2021 19:26:03 +0000 (14:26 -0500)]
NFSD: De-duplicate net_generic(nf->nf_net, nfsd_net_id)

[ Upstream commit 2c445a0e72cb1fbfbdb7f9473c53556ee27c1d90 ]

Since this pointer is used repeatedly, move it to a stack variable.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agodrm/amd/display: Fix by adding FPU protection for dcn30_internal_validate_bw
CHANDAN VURDIGERE NATARAJ [Tue, 29 Mar 2022 07:40:31 +0000 (13:10 +0530)]
drm/amd/display: Fix by adding FPU protection for dcn30_internal_validate_bw

[ Upstream commit 50e6cb3fd2cde554db646282ea10df7236e6493c ]

[Why]
Below general protection fault observed when WebGL Aquarium is run for
longer duration. If drm debug logs are enabled and set to 0x1f then the
issue is observed within 10 minutes of run.

[  100.717056] general protection fault, probably for non-canonical address 0x2d33302d32323032: 0000 [#1] PREEMPT SMP NOPTI
[  100.727921] CPU: 3 PID: 1906 Comm: DrmThread Tainted: G        W         5.15.30 #12 d726c6a2d6ebe5cf9223931cbca6892f916fe18b
[  100.754419] RIP: 0010:CalculateSwathWidth+0x1f7/0x44f
[  100.767109] Code: 00 00 00 f2 42 0f 11 04 f0 48 8b 85 88 00 00 00 f2 42 0f 10 04 f0 48 8b 85 98 00 00 00 f2 42 0f 11 04 f0 48 8b 45 10 0f 57 c0 <f3> 42 0f 2a 04 b0 0f 57 c9 f3 43 0f 2a 0c b4 e8 8c e2 f3 ff 48 8b
[  100.781269] RSP: 0018:ffffa9230079eeb0 EFLAGS: 00010246
[  100.812528] RAX: 2d33302d32323032 RBX: 0000000000000500 RCX: 0000000000000000
[  100.819656] RDX: 0000000000000001 RSI: ffff99deb712c49c RDI: 0000000000000000
[  100.826781] RBP: ffffa9230079ef50 R08: ffff99deb712460c R09: ffff99deb712462c
[  100.833907] R10: ffff99deb7124940 R11: ffff99deb7124d70 R12: ffff99deb712ae44
[  100.841033] R13: 0000000000000001 R14: 0000000000000000 R15: ffffa9230079f0a0
[  100.848159] FS:  00007af121212640(0000) GS:ffff99deba780000(0000) knlGS:0000000000000000
[  100.856240] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  100.861980] CR2: 0000209000fe1000 CR3: 000000011b18c000 CR4: 0000000000350ee0
[  100.869106] Call Trace:
[  100.871555]  <TASK>
[  100.873655]  ? asm_sysvec_reschedule_ipi+0x12/0x20
[  100.878449]  CalculateSwathAndDETConfiguration+0x1a3/0x6dd
[  100.883937]  dml31_ModeSupportAndSystemConfigurationFull+0x2ce4/0x76da
[  100.890467]  ? kallsyms_lookup_buildid+0xc8/0x163
[  100.895173]  ? kallsyms_lookup_buildid+0xc8/0x163
[  100.899874]  ? __sprint_symbol+0x80/0x135
[  100.903883]  ? dm_update_plane_state+0x3f9/0x4d2
[  100.908500]  ? symbol_string+0xb7/0xde
[  100.912250]  ? number+0x145/0x29b
[  100.915566]  ? vsnprintf+0x341/0x5ff
[  100.919141]  ? desc_read_finalized_seq+0x39/0x87
[  100.923755]  ? update_load_avg+0x1b9/0x607
[  100.927849]  ? compute_mst_dsc_configs_for_state+0x7d/0xd5b
[  100.933416]  ? fetch_pipe_params+0xa4d/0xd0c
[  100.937686]  ? dc_fpu_end+0x3d/0xa8
[  100.941175]  dml_get_voltage_level+0x16b/0x180
[  100.945619]  dcn30_internal_validate_bw+0x10e/0x89b
[  100.950495]  ? dcn31_validate_bandwidth+0x68/0x1fc
[  100.955285]  ? resource_build_scaling_params+0x98b/0xb8c
[  100.960595]  ? dcn31_validate_bandwidth+0x68/0x1fc
[  100.965384]  dcn31_validate_bandwidth+0x9a/0x1fc
[  100.970001]  dc_validate_global_state+0x238/0x295
[  100.974703]  amdgpu_dm_atomic_check+0x9c1/0xbce
[  100.979235]  ? _printk+0x59/0x73
[  100.982467]  drm_atomic_check_only+0x403/0x78b
[  100.986912]  drm_mode_atomic_ioctl+0x49b/0x546
[  100.991358]  ? drm_ioctl+0x1c1/0x3b3
[  100.994936]  ? drm_atomic_set_property+0x92a/0x92a
[  100.999725]  drm_ioctl_kernel+0xdc/0x149
[  101.003648]  drm_ioctl+0x27f/0x3b3
[  101.007051]  ? drm_atomic_set_property+0x92a/0x92a
[  101.011842]  amdgpu_drm_ioctl+0x49/0x7d
[  101.015679]  __se_sys_ioctl+0x7c/0xb8
[  101.015685]  do_syscall_64+0x5f/0xb8
[  101.015690]  ? __irq_exit_rcu+0x34/0x96

[How]
It calles populate_dml_pipes which uses doubles to initialize.
Adding FPU protection avoids context switch and probable loss of vba context
as there is potential contention while drm debug logs are enabled.

Signed-off-by: CHANDAN VURDIGERE NATARAJ <chandan.vurdigerenataraj@amd.com>
Reviewed-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agodrm/amd/display: Set min dcfclk if pipe count is 0
Michael Strauss [Wed, 11 Aug 2021 15:38:44 +0000 (11:38 -0400)]
drm/amd/display: Set min dcfclk if pipe count is 0

[ Upstream commit bc204778b4032b336cb3bde85bea852d79e7e389 ]

[WHY]
Clocks don't get recalculated in 0 stream/0 pipe configs,
blocking S0i3 if dcfclk gets high enough

[HOW]
Create DCN31 copy of DCN30 bandwidth validation func which
doesn't entirely skip validation in 0 pipe scenarios

Override dcfclk to vlevel 0/min value during validation if pipe
count is 0

Reviewed-by: Eric Yang <Eric.Yang2@amd.com>
Acked-by: Qingqing Zhuo <qingqing.zhuo@amd.com>
Signed-off-by: Michael Strauss <michael.strauss@amd.com>
Tested-by: Daniel Wheeler <Daniel.Wheeler@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agodrbd: fix an invalid memory access caused by incorrect use of list iterator
Xiaomeng Tong [Wed, 6 Apr 2022 19:04:44 +0000 (21:04 +0200)]
drbd: fix an invalid memory access caused by incorrect use of list iterator

[ Upstream commit ae4d37b5df749926891583d42a6801b5da11e3c1 ]

The bug is here:
idr_remove(&connection->peer_devices, vnr);

If the previous for_each_connection() don't exit early (no goto hit
inside the loop), the iterator 'connection' after the loop will be a
bogus pointer to an invalid structure object containing the HEAD
(&resource->connections). As a result, the use of 'connection' above
will lead to a invalid memory access (including a possible invalid free
as idr_remove could call free_layer).

The original intention should have been to remove all peer_devices,
but the following lines have already done the work. So just remove
this line and the unneeded label, to fix this bug.

Cc: stable@vger.kernel.org
Fixes: c06ece6ba6f1b ("drbd: Turn connection->volumes into connection->peer_devices")
Signed-off-by: Xiaomeng Tong <xiam0nd.tong@gmail.com>
Reviewed-by: Christoph Böhmwalder <christoph.boehmwalder@linbit.com>
Reviewed-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agodrbd: Fix double free problem in drbd_create_device
Wu Bo [Thu, 4 Nov 2021 08:07:09 +0000 (16:07 +0800)]
drbd: Fix double free problem in drbd_create_device

[ Upstream commit 27548088ac628109f70eb0b1eb521d035844dba8 ]

In drbd_create_device(), the 'out_no_io_page' lable has called
blk_cleanup_disk() when return failed.

So remove the 'out_cleanup_disk' lable to avoid double free the
disk pointer.

Fixes: e92ab4eda516 ("drbd: add error handling support for add_disk()")
Signed-off-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/1636013229-26309-1-git-send-email-wubo40@huawei.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agodrbd: add error handling support for add_disk()
Luis Chamberlain [Mon, 27 Sep 2021 22:00:59 +0000 (15:00 -0700)]
drbd: add error handling support for add_disk()

[ Upstream commit e92ab4eda516a5bfd96c087282ebe9521deba4f4 ]

We never checked for errors on add_disk() as this function
returned void. Now that this is fixed, use the shiny new
error handling.

Signed-off-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: remove device item and update super block in the same transaction
Qu Wenruo [Tue, 8 Mar 2022 05:36:38 +0000 (13:36 +0800)]
btrfs: remove device item and update super block in the same transaction

[ Upstream commit bbac58698a55cc0a6f0c0d69a6dcd3f9f3134c11 ]

[BUG]
There is a report that a btrfs has a bad super block num devices.

This makes btrfs to reject the fs completely.

  BTRFS error (device sdd3): super_num_devices 3 mismatch with num_devices 2 found here
  BTRFS error (device sdd3): failed to read chunk tree: -22
  BTRFS error (device sdd3): open_ctree failed

[CAUSE]
During btrfs device removal, chunk tree and super block num devs are
updated in two different transactions:

  btrfs_rm_device()
  |- btrfs_rm_dev_item(device)
  |  |- trans = btrfs_start_transaction()
  |  |  Now we got transaction X
  |  |
  |  |- btrfs_del_item()
  |  |  Now device item is removed from chunk tree
  |  |
  |  |- btrfs_commit_transaction()
  |     Transaction X got committed, super num devs untouched,
  |     but device item removed from chunk tree.
  |     (AKA, super num devs is already incorrect)
  |
  |- cur_devices->num_devices--;
  |- cur_devices->total_devices--;
  |- btrfs_set_super_num_devices()
     All those operations are not in transaction X, thus it will
     only be written back to disk in next transaction.

So after the transaction X in btrfs_rm_dev_item() committed, but before
transaction X+1 (which can be minutes away), a power loss happen, then
we got the super num mismatch.

[FIX]
Instead of starting and committing a transaction inside
btrfs_rm_dev_item(), start a transaction in side btrfs_rm_device() and
pass it to btrfs_rm_dev_item().

And only commit the transaction after everything is done.

Reported-by: Luca Béla Palkovics <luca.bela.palkovics@gmail.com>
Link: https://lore.kernel.org/linux-btrfs/CA+8xDSpvdm_U0QLBAnrH=zqDq_cWCOH5TiV46CKmp3igr44okQ@mail.gmail.com/
CC: stable@vger.kernel.org # 4.14+
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: use btrfs_get_dev_args_from_path in dev removal ioctls
Josef Bacik [Tue, 5 Oct 2021 20:12:44 +0000 (16:12 -0400)]
btrfs: use btrfs_get_dev_args_from_path in dev removal ioctls

[ Upstream commit 1a15eb724aaef8656f8cc01d9355797cfe7c618e ]

For device removal and replace we call btrfs_find_device_by_devspec,
which if we give it a device path and nothing else will call
btrfs_get_dev_args_from_path, which opens the block device and reads the
super block and then looks up our device based on that.

However at this point we're holding the sb write "lock", so reading the
block device pulls in the dependency of ->open_mutex, which produces the
following lockdep splat

======================================================
WARNING: possible circular locking dependency detected
5.14.0-rc2+ #405 Not tainted
------------------------------------------------------
losetup/11576 is trying to acquire lock:
ffff9bbe8cded938 ((wq_completion)loop0){+.+.}-{0:0}, at: flush_workqueue+0x67/0x5e0

but task is already holding lock:
ffff9bbe88e4fc68 (&lo->lo_mutex){+.+.}-{3:3}, at: __loop_clr_fd+0x41/0x660 [loop]

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #4 (&lo->lo_mutex){+.+.}-{3:3}:
       __mutex_lock+0x7d/0x750
       lo_open+0x28/0x60 [loop]
       blkdev_get_whole+0x25/0xf0
       blkdev_get_by_dev.part.0+0x168/0x3c0
       blkdev_open+0xd2/0xe0
       do_dentry_open+0x161/0x390
       path_openat+0x3cc/0xa20
       do_filp_open+0x96/0x120
       do_sys_openat2+0x7b/0x130
       __x64_sys_openat+0x46/0x70
       do_syscall_64+0x38/0x90
       entry_SYSCALL_64_after_hwframe+0x44/0xae

-> #3 (&disk->open_mutex){+.+.}-{3:3}:
       __mutex_lock+0x7d/0x750
       blkdev_get_by_dev.part.0+0x56/0x3c0
       blkdev_get_by_path+0x98/0xa0
       btrfs_get_bdev_and_sb+0x1b/0xb0
       btrfs_find_device_by_devspec+0x12b/0x1c0
       btrfs_rm_device+0x127/0x610
       btrfs_ioctl+0x2a31/0x2e70
       __x64_sys_ioctl+0x80/0xb0
       do_syscall_64+0x38/0x90
       entry_SYSCALL_64_after_hwframe+0x44/0xae

-> #2 (sb_writers#12){.+.+}-{0:0}:
       lo_write_bvec+0xc2/0x240 [loop]
       loop_process_work+0x238/0xd00 [loop]
       process_one_work+0x26b/0x560
       worker_thread+0x55/0x3c0
       kthread+0x140/0x160
       ret_from_fork+0x1f/0x30

-> #1 ((work_completion)(&lo->rootcg_work)){+.+.}-{0:0}:
       process_one_work+0x245/0x560
       worker_thread+0x55/0x3c0
       kthread+0x140/0x160
       ret_from_fork+0x1f/0x30

-> #0 ((wq_completion)loop0){+.+.}-{0:0}:
       __lock_acquire+0x10ea/0x1d90
       lock_acquire+0xb5/0x2b0
       flush_workqueue+0x91/0x5e0
       drain_workqueue+0xa0/0x110
       destroy_workqueue+0x36/0x250
       __loop_clr_fd+0x9a/0x660 [loop]
       block_ioctl+0x3f/0x50
       __x64_sys_ioctl+0x80/0xb0
       do_syscall_64+0x38/0x90
       entry_SYSCALL_64_after_hwframe+0x44/0xae

other info that might help us debug this:

Chain exists of:
  (wq_completion)loop0 --> &disk->open_mutex --> &lo->lo_mutex

 Possible unsafe locking scenario:

       CPU0                    CPU1
       ----                    ----
  lock(&lo->lo_mutex);
                               lock(&disk->open_mutex);
                               lock(&lo->lo_mutex);
  lock((wq_completion)loop0);

 *** DEADLOCK ***

1 lock held by losetup/11576:
 #0: ffff9bbe88e4fc68 (&lo->lo_mutex){+.+.}-{3:3}, at: __loop_clr_fd+0x41/0x660 [loop]

stack backtrace:
CPU: 0 PID: 11576 Comm: losetup Not tainted 5.14.0-rc2+ #405
Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 1.13.0-2.fc32 04/01/2014
Call Trace:
 dump_stack_lvl+0x57/0x72
 check_noncircular+0xcf/0xf0
 ? stack_trace_save+0x3b/0x50
 __lock_acquire+0x10ea/0x1d90
 lock_acquire+0xb5/0x2b0
 ? flush_workqueue+0x67/0x5e0
 ? lockdep_init_map_type+0x47/0x220
 flush_workqueue+0x91/0x5e0
 ? flush_workqueue+0x67/0x5e0
 ? verify_cpu+0xf0/0x100
 drain_workqueue+0xa0/0x110
 destroy_workqueue+0x36/0x250
 __loop_clr_fd+0x9a/0x660 [loop]
 ? blkdev_ioctl+0x8d/0x2a0
 block_ioctl+0x3f/0x50
 __x64_sys_ioctl+0x80/0xb0
 do_syscall_64+0x38/0x90
 entry_SYSCALL_64_after_hwframe+0x44/0xae
RIP: 0033:0x7f31b02404cb

Instead what we want to do is populate our device lookup args before we
grab any locks, and then pass these args into btrfs_rm_device().  From
there we can find the device and do the appropriate removal.

Suggested-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: add a btrfs_get_dev_args_from_path helper
Josef Bacik [Tue, 5 Oct 2021 20:12:43 +0000 (16:12 -0400)]
btrfs: add a btrfs_get_dev_args_from_path helper

[ Upstream commit faa775c41d655a4786e9d53cb075a77bb5a75f66 ]

We are going to want to populate our device lookup args outside of any
locks and then do the actual device lookup later, so add a helper to do
this work and make btrfs_find_device_by_devspec() use this helper for
now.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: handle device lookup with btrfs_dev_lookup_args
Josef Bacik [Tue, 5 Oct 2021 20:12:42 +0000 (16:12 -0400)]
btrfs: handle device lookup with btrfs_dev_lookup_args

[ Upstream commit 562d7b1512f7369a19bca2883e2e8672d78f0481 ]

We have a lot of device lookup functions that all do something slightly
different.  Clean this up by adding a struct to hold the different
lookup criteria, and then pass this around to btrfs_find_device() so it
can do the proper matching based on the lookup criteria.

Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agovdpa/mlx5: Avoid processing works if workqueue was destroyed
Eli Cohen [Mon, 21 Mar 2022 14:13:03 +0000 (16:13 +0200)]
vdpa/mlx5: Avoid processing works if workqueue was destroyed

[ Upstream commit ad6dc1daaf29f97f23cc810d60ee01c0e83f4c6b ]

If mlx5_vdpa gets unloaded while a VM is running, the workqueue will be
destroyed. However, vhost might still have reference to the kick
function and might attempt to push new works. This could lead to null
pointer dereference.

To fix this, set mvdev->wq to NULL just before destroying and verify
that the workqueue is not NULL in mlx5_vdpa_kick_vq before attempting to
push a new work.

Fixes: 5262912ef3cf ("vdpa/mlx5: Add support for control VQ and MAC setting")
Signed-off-by: Eli Cohen <elic@nvidia.com>
Link: https://lore.kernel.org/r/20220321141303.9586-1-elic@nvidia.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agogfs2: Fix gfs2_file_buffered_write endless loop workaround
Andreas Gruenbacher [Thu, 17 Mar 2022 13:47:24 +0000 (14:47 +0100)]
gfs2: Fix gfs2_file_buffered_write endless loop workaround

[ Upstream commit 46f3e0421ccb5474b5c006b0089b9dfd42534bb6 ]

Since commit 554c577cee95b, gfs2_file_buffered_write() can accidentally
return a truncated iov_iter, which might confuse callers.  Fix that.

Fixes: 554c577cee95b ("gfs2: Prevent endless loops in gfs2_file_buffered_write")
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoscsi: qla2xxx: Fix crash during module load unload test
Arun Easi [Thu, 10 Mar 2022 09:25:56 +0000 (01:25 -0800)]
scsi: qla2xxx: Fix crash during module load unload test

[ Upstream commit 0972252450f90db56dd5415a20e2aec21a08d036 ]

During purex packet handling the driver was incorrectly freeing a
pre-allocated structure. Fix this by skipping that entry.

System crashed with the following stack during a module unload test.

Call Trace:
sbitmap_init_node+0x7f/0x1e0
sbitmap_queue_init_node+0x24/0x150
blk_mq_init_bitmaps+0x3d/0xa0
blk_mq_init_tags+0x68/0x90
blk_mq_alloc_map_and_rqs+0x44/0x120
blk_mq_alloc_set_map_and_rqs+0x63/0x150
blk_mq_alloc_tag_set+0x11b/0x230
scsi_add_host_with_dma.cold+0x3f/0x245
qla2x00_probe_one+0xd5a/0x1b80 [qla2xxx]

Call Trace with slub_debug and debug kernel:
kasan_report_invalid_free+0x50/0x80
__kasan_slab_free+0x137/0x150
slab_free_freelist_hook+0xc6/0x190
kfree+0xe8/0x2e0
qla2x00_free_device+0x3bb/0x5d0 [qla2xxx]
qla2x00_remove_one+0x668/0xcf0 [qla2xxx]

Link: https://lore.kernel.org/r/20220310092604.22950-6-njavali@marvell.com
Fixes: 62e9dd177732 ("scsi: qla2xxx: Change in PUREX to handle FPIN ELS requests")
Cc: stable@vger.kernel.org
Reported-by: Marco Patalano <mpatalan@redhat.com>
Tested-by: Marco Patalano <mpatalan@redhat.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoscsi: qla2xxx: edif: Replace list_for_each_safe with list_for_each_entry_safe
Quinn Tran [Tue, 26 Oct 2021 11:54:06 +0000 (04:54 -0700)]
scsi: qla2xxx: edif: Replace list_for_each_safe with list_for_each_entry_safe

[ Upstream commit 8062b742d3bd336ca10ab5a1db1629d33700f9c6 ]

This patch is per review comment by Hannes Reinecke from previous
submission to replace list_for_each_safe with list_for_each_entry_safe.

Link: https://lore.kernel.org/r/20211026115412.27691-8-njavali@marvell.com
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoscsi: qla2xxx: Fix laggy FC remote port session recovery
Quinn Tran [Thu, 10 Mar 2022 09:25:59 +0000 (01:25 -0800)]
scsi: qla2xxx: Fix laggy FC remote port session recovery

[ Upstream commit 713b415726f100f6644971e75ebfe1edbef1a390 ]

For session recovery, driver relies on the dpc thread to initiate certain
operations. The dpc thread runs exclusively without the Mailbox interface
being occupied. A recent code change for heartbeat check via mailbox cmd 0
is preventing the dpc thread from carrying out its operation. This patch
allows the higher priority error recovery to run first before running the
lower priority heartbeat check.

Link: https://lore.kernel.org/r/20220310092604.22950-9-njavali@marvell.com
Fixes: d94d8158e184 ("scsi: qla2xxx: Add heartbeat check")
Cc: stable@vger.kernel.org
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoscsi: qla2xxx: Move heartbeat handling from DPC thread to workqueue
Manish Rangankar [Wed, 8 Sep 2021 16:46:20 +0000 (09:46 -0700)]
scsi: qla2xxx: Move heartbeat handling from DPC thread to workqueue

[ Upstream commit 3a4e1f3b3a3c733de3b82b9b522e54803e1165ae ]

DPC thread gets restricted due to a no-op mailbox, which is a blocking call
and has a high execution frequency. To free up the DPC thread we move no-op
handling to the workqueue. Also, modified qla_do_heartbeat() to send no-op
MBC if we don’t have any active interrupts, but there are still I/Os
outstanding with firmware.

Link: https://lore.kernel.org/r/20210908164622.19240-9-njavali@marvell.com
Fixes: d94d8158e184 ("scsi: qla2xxx: Add heartbeat check")
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Manish Rangankar <mrangankar@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoKVM: x86/mmu: Use common TDP MMU zap helper for MMU notifier unmap hook
Sean Christopherson [Wed, 15 Dec 2021 01:15:54 +0000 (01:15 +0000)]
KVM: x86/mmu: Use common TDP MMU zap helper for MMU notifier unmap hook

[ Upstream commit 83b83a02073ec8d18c77a9bbe0881d710f7a9d32 ]

Use the common TDP MMU zap helper when handling an MMU notifier unmap
event, the two flows are semantically identical.  Consolidate the code in
preparation for a future bug fix, as both kvm_tdp_mmu_unmap_gfn_range()
and __kvm_tdp_mmu_zap_gfn_range() are guilty of not zapping SPTEs in
invalid roots.

No functional change intended.

Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20211215011557.399940-2-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoKVM: x86/mmu: Use yield-safe TDP MMU root iter in MMU notifier unmapping
Sean Christopherson [Sat, 20 Nov 2021 01:50:08 +0000 (01:50 +0000)]
KVM: x86/mmu: Use yield-safe TDP MMU root iter in MMU notifier unmapping

[ Upstream commit 7533377215b6ee432c06c5855f6be5d66e694e46 ]

Use the yield-safe variant of the TDP MMU iterator when handling an
unmapping event from the MMU notifier, as most occurences of the event
allow yielding.

Fixes: e1eed5847b09 ("KVM: x86/mmu: Allow yielding during MMU notifier unmap/zap, if possible")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20211120015008.3780032-1-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoclk: renesas: r9a07g044: Update multiplier and divider values for PLL2/3
Lad Prabhakar [Thu, 23 Dec 2021 09:32:23 +0000 (09:32 +0000)]
clk: renesas: r9a07g044: Update multiplier and divider values for PLL2/3

[ Upstream commit b289cdecc7c3e25e001cde260c882e4d9a8b0772 ]

As per the HW manual (Rev.1.00 Sep, 2021) PLL2 and PLL3 should be
1600 MHz, but with current multiplier and divider values this resulted
to 1596 MHz.

This patch updates the multiplier and divider values for PLL2 and PLL3
so that we get the exact (1600 MHz) values.

Fixes: 17f0ff3d49ff1 ("clk: renesas: Add support for R9A07G044 SoC")
Suggested-by: Biju Das <biju.das.jz@bp.renesas.com>
Signed-off-by: Lad Prabhakar <prabhakar.mahadev-lad.rj@bp.renesas.com>
Link: https://lore.kernel.org/r/20211223093223.4725-1-prabhakar.mahadev-lad.rj@bp.renesas.com
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agocxl/port: Hold port reference until decoder release
Dan Williams [Thu, 17 Feb 2022 00:25:11 +0000 (16:25 -0800)]
cxl/port: Hold port reference until decoder release

[ Upstream commit 74be98774dfbc5b8b795db726bd772e735d2edd4 ]

KASAN + DEBUG_KOBJECT_RELEASE reports a potential use-after-free in
cxl_decoder_release() where it goes to reference its parent, a cxl_port,
to free its id back to port->decoder_ida.

 BUG: KASAN: use-after-free in to_cxl_port+0x18/0x90 [cxl_core]
 Read of size 8 at addr ffff888119270908 by task kworker/35:2/379

 CPU: 35 PID: 379 Comm: kworker/35:2 Tainted: G           OE     5.17.0-rc2+ #198
 Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 0.0.0 02/06/2015
 Workqueue: events kobject_delayed_cleanup
 Call Trace:
  <TASK>
  dump_stack_lvl+0x59/0x73
  print_address_description.constprop.0+0x1f/0x150
  ? to_cxl_port+0x18/0x90 [cxl_core]
  kasan_report.cold+0x83/0xdf
  ? to_cxl_port+0x18/0x90 [cxl_core]
  to_cxl_port+0x18/0x90 [cxl_core]
  cxl_decoder_release+0x2a/0x60 [cxl_core]
  device_release+0x5f/0x100
  kobject_cleanup+0x80/0x1c0

The device core only guarantees parent lifetime until all children are
unregistered. If a child needs a parent to complete its ->release()
callback that child needs to hold a reference to extend the lifetime of
the parent.

Fixes: 40ba17afdfab ("cxl/acpi: Introduce cxl_decoder objects")
Reported-by: Ben Widawsky <ben.widawsky@intel.com>
Tested-by: Ben Widawsky <ben.widawsky@intel.com>
Reviewed-by: Ben Widawsky <ben.widawsky@intel.com>
Link: https://lore.kernel.org/r/164505751190.4175768.13324905271463416712.stgit@dwillia2-desk3.amr.corp.intel.com
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomt76: mt7921: do not always disable fw runtime-pm
Lorenzo Bianconi [Fri, 31 Dec 2021 11:36:02 +0000 (12:36 +0100)]
mt76: mt7921: do not always disable fw runtime-pm

[ Upstream commit b44eeb8cbdf2b88f2844f11e4f263b0abed5b5b0 ]

After commit 'd430dffbe9dd ("mt76: mt7921: fix a possible race
enabling/disabling runtime-pm")', runtime-pm is always disabled in the
fw even if the user requests to enable it toggling debugfs node since
mt7921_pm_interface_iter routine will use pm->enable to configure the fw.
Fix the issue moving enable variable configuration before running
mt7921_pm_interface_iter routine.

Fixes: d430dffbe9dd ("mt76: mt7921: fix a possible race enabling/disabling runtime-pm")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomt76: mt76_connac: fix MCU_CE_CMD_SET_ROC definition error
Sean Wang [Fri, 24 Dec 2021 08:33:55 +0000 (16:33 +0800)]
mt76: mt76_connac: fix MCU_CE_CMD_SET_ROC definition error

[ Upstream commit bf9727a27442a50c75b7d99a5088330c578b2a42 ]

Fixed an MCU_CE_CMD_SET_ROC definition error that occurred from a previous
refactor work.

Fixes: d0e274af2f2e4 ("mt76: mt76_connac: create mcu library")
Signed-off-by: Sean Wang <sean.wang@mediatek.com>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomedia: davinci: vpif: fix use-after-free on driver unbind
Johan Hovold [Wed, 22 Dec 2021 14:20:24 +0000 (15:20 +0100)]
media: davinci: vpif: fix use-after-free on driver unbind

[ Upstream commit 43acb728bbc40169d2e2425e84a80068270974be ]

The driver allocates and registers two platform device structures during
probe, but the devices were never deregistered on driver unbind.

This results in a use-after-free on driver unbind as the device
structures were allocated using devres and would be freed by driver
core when remove() returns.

Fix this by adding the missing deregistration calls to the remove()
callback and failing probe on registration errors.

Note that the platform device structures must be freed using a proper
release callback to avoid leaking associated resources like device
names.

Fixes: 479f7a118105 ("[media] davinci: vpif: adaptions for DT support")
Cc: stable@vger.kernel.org # 4.12
Cc: Kevin Hilman <khilman@baylibre.com>
Signed-off-by: Johan Hovold <johan@kernel.org>
Reviewed-by: Lad Prabhakar <prabhakar.csengg@gmail.com>
Signed-off-by: Hans Verkuil <hverkuil-cisco@xs4all.nl>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomedia: omap3isp: Use struct_group() for memcpy() region
Kees Cook [Mon, 24 Jan 2022 17:29:52 +0000 (18:29 +0100)]
media: omap3isp: Use struct_group() for memcpy() region

[ Upstream commit d4568fc8525897e683983806f813be1ae9eedaed ]

In preparation for FORTIFY_SOURCE performing compile-time and run-time
field bounds checking for memcpy(), memmove(), and memset(), avoid
intentionally writing across neighboring fields. Wrap the target region
in struct_group(). This additionally fixes a theoretical misalignment
of the copy (since the size of "buf" changes between 64-bit and 32-bit,
but this is likely never built for 64-bit).

FWIW, I think this code is totally broken on 64-bit (which appears to
not be a "real" build configuration): it would either always fail (with
an uninitialized data->buf_size) or would cause corruption in userspace
due to the copy_to_user() in the call path against an uninitialized
data->buf value:

omap3isp_stat_request_statistics_time32(...)
    struct omap3isp_stat_data data64;
    ...
    omap3isp_stat_request_statistics(stat, &data64);

int omap3isp_stat_request_statistics(struct ispstat *stat,
                                     struct omap3isp_stat_data *data)
    ...
    buf = isp_stat_buf_get(stat, data);

static struct ispstat_buffer *isp_stat_buf_get(struct ispstat *stat,
                                               struct omap3isp_stat_data *data)
...
    if (buf->buf_size > data->buf_size) {
            ...
            return ERR_PTR(-EINVAL);
    }
    ...
    rval = copy_to_user(data->buf,
                        buf->virt_addr,
                        buf->buf_size);

Regardless, additionally initialize data64 to be zero-filled to avoid
undefined behavior.

Link: https://lore.kernel.org/lkml/20211215220505.GB21862@embeddedor
Cc: Arnd Bergmann <arnd@arndb.de>
Fixes: 378e3f81cb56 ("media: omap3isp: support 64-bit version of omap3isp_stat_data")
Cc: stable@vger.kernel.org
Reviewed-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Laurent Pinchart <laurent.pinchart@ideasonboard.com>
Signed-off-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agostddef: Introduce struct_group() helper macro
Kees Cook [Tue, 18 May 2021 03:01:15 +0000 (20:01 -0700)]
stddef: Introduce struct_group() helper macro

[ Upstream commit 50d7bd38c3aafc4749e05e8d7fcb616979143602 ]

Kernel code has a regular need to describe groups of members within a
structure usually when they need to be copied or initialized separately
from the rest of the surrounding structure. The generally accepted design
pattern in C is to use a named sub-struct:

struct foo {
int one;
struct {
int two;
int three, four;
} thing;
int five;
};

This would allow for traditional references and sizing:

memcpy(&dst.thing, &src.thing, sizeof(dst.thing));

However, doing this would mean that referencing struct members enclosed
by such named structs would always require including the sub-struct name
in identifiers:

do_something(dst.thing.three);

This has tended to be quite inflexible, especially when such groupings
need to be added to established code which causes huge naming churn.
Three workarounds exist in the kernel for this problem, and each have
other negative properties.

To avoid the naming churn, there is a design pattern of adding macro
aliases for the named struct:

#define f_three thing.three

This ends up polluting the global namespace, and makes it difficult to
search for identifiers.

Another common work-around in kernel code avoids the pollution by avoiding
the named struct entirely, instead identifying the group's boundaries using
either a pair of empty anonymous structs of a pair of zero-element arrays:

struct foo {
int one;
struct { } start;
int two;
int three, four;
struct { } finish;
int five;
};

struct foo {
int one;
int start[0];
int two;
int three, four;
int finish[0];
int five;
};

This allows code to avoid needing to use a sub-struct named for member
references within the surrounding structure, but loses the benefits of
being able to actually use such a struct, making it rather fragile. Using
these requires open-coded calculation of sizes and offsets. The efforts
made to avoid common mistakes include lots of comments, or adding various
BUILD_BUG_ON()s. Such code is left with no way for the compiler to reason
about the boundaries (e.g. the "start" object looks like it's 0 bytes
in length), making bounds checking depend on open-coded calculations:

if (length > offsetof(struct foo, finish) -
     offsetof(struct foo, start))
return -EINVAL;
memcpy(&dst.start, &src.start, offsetof(struct foo, finish) -
       offsetof(struct foo, start));

However, the vast majority of places in the kernel that operate on
groups of members do so without any identification of the grouping,
relying either on comments or implicit knowledge of the struct contents,
which is even harder for the compiler to reason about, and results in
even more fragile manual sizing, usually depending on member locations
outside of the region (e.g. to copy "two" and "three", use the start of
"four" to find the size):

BUILD_BUG_ON((offsetof(struct foo, four) <
      offsetof(struct foo, two)) ||
     (offsetof(struct foo, four) <
      offsetof(struct foo, three));
if (length > offsetof(struct foo, four) -
     offsetof(struct foo, two))
return -EINVAL;
memcpy(&dst.two, &src.two, length);

In order to have a regular programmatic way to describe a struct
region that can be used for references and sizing, can be examined for
bounds checking, avoids forcing the use of intermediate identifiers,
and avoids polluting the global namespace, introduce the struct_group()
macro. This macro wraps the member declarations to create an anonymous
union of an anonymous struct (no intermediate name) and a named struct
(for references and sizing):

struct foo {
int one;
struct_group(thing,
int two;
int three, four;
);
int five;
};

if (length > sizeof(src.thing))
return -EINVAL;
memcpy(&dst.thing, &src.thing, length);
do_something(dst.three);

There are some rare cases where the resulting struct_group() needs
attributes added, so struct_group_attr() is also introduced to allow
for specifying struct attributes (e.g. __align(x) or __packed).
Additionally, there are places where such declarations would like to
have the struct be tagged, so struct_group_tagged() is added.

Given there is a need for a handful of UAPI uses too, the underlying
__struct_group() macro has been defined in UAPI so it can be used there
too.

To avoid confusing scripts/kernel-doc, hide the macro from its struct
parsing.

Co-developed-by: Keith Packard <keithp@keithp.com>
Signed-off-by: Keith Packard <keithp@keithp.com>
Acked-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Link: https://lore.kernel.org/lkml/20210728023217.GC35706@embeddedor
Enhanced-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Link: https://lore.kernel.org/lkml/41183a98-bdb9-4ad6-7eab-5a7292a6df84@rasmusvillemoes.dk
Enhanced-by: Dan Williams <dan.j.williams@intel.com>
Link: https://lore.kernel.org/lkml/1d9a2e6df2a9a35b2cdd50a9a68cac5991e7e5f0.camel@intel.com
Enhanced-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Link: https://lore.kernel.org/lkml/YQKa76A6XuFqgM03@phenom.ffwll.local
Acked-by: Dan Williams <dan.j.williams@intel.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoblock: fix rq-qos breakage from skipping rq_qos_done_bio()
Tejun Heo [Mon, 14 Mar 2022 07:15:02 +0000 (21:15 -1000)]
block: fix rq-qos breakage from skipping rq_qos_done_bio()

[ Upstream commit aa1b46dcdc7baaf5fec0be25782ef24b26aa209e ]

a647a524a467 ("block: don't call rq_qos_ops->done_bio if the bio isn't
tracked") made bio_endio() skip rq_qos_done_bio() if BIO_TRACKED is not set.
While this fixed a potential oops, it also broke blk-iocost by skipping the
done_bio callback for merged bios.

Before, whether a bio goes through rq_qos_throttle() or rq_qos_merge(),
rq_qos_done_bio() would be called on the bio on completion with BIO_TRACKED
distinguishing the former from the latter. rq_qos_done_bio() is not called
for bios which wenth through rq_qos_merge(). This royally confuses
blk-iocost as the merged bios never finish and are considered perpetually
in-flight.

One reliably reproducible failure mode is an intermediate cgroup geting
stuck active preventing its children from being activated due to the
leaf-only rule, leading to loss of control. The following is from
resctl-bench protection scenario which emulates isolating a web server like
workload from a memory bomb run on an iocost configuration which should
yield a reasonable level of protection.

  # cat /sys/block/nvme2n1/device/model
  Samsung SSD 970 PRO 512GB
  # cat /sys/fs/cgroup/io.cost.model
  259:0 ctrl=user model=linear rbps=834913556 rseqiops=93622 rrandiops=102913 wbps=618985353 wseqiops=72325 wrandiops=71025
  # cat /sys/fs/cgroup/io.cost.qos
  259:0 enable=1 ctrl=user rpct=95.00 rlat=18776 wpct=95.00 wlat=8897 min=60.00 max=100.00
  # resctl-bench -m 29.6G -r out.json run protection::scenario=mem-hog,loops=1
  ...
  Memory Hog Summary
  ==================

  IO Latency: R p50=242u:336u/2.5m p90=794u:1.4m/7.5m p99=2.7m:8.0m/62.5m max=8.0m:36.4m/350m
              W p50=221u:323u/1.5m p90=709u:1.2m/5.5m p99=1.5m:2.5m/9.5m max=6.9m:35.9m/350m

  Isolation and Request Latency Impact Distributions:

                min   p01   p05   p10   p25   p50   p75   p90   p95   p99   max  mean stdev
  isol%       15.90 15.90 15.90 40.05 57.24 59.07 60.01 74.63 74.63 90.35 90.35 58.12 15.82
  lat-imp%        0     0     0     0     0  4.55 14.68 15.54 233.5 548.1 548.1 53.88 143.6

  Result: isol=58.12:15.82% lat_imp=53.88%:143.6 work_csv=100.0% missing=3.96%

The isolation result of 58.12% is close to what this device would show
without any IO control.

Fix it by introducing a new flag BIO_QOS_MERGED to mark merged bios and
calling rq_qos_done_bio() on them too. For consistency and clarity, rename
BIO_TRACKED to BIO_QOS_THROTTLED. The flag checks are moved into
rq_qos_done_bio() so that it's next to the code paths that set the flags.

With the patch applied, the above same benchmark shows:

  # resctl-bench -m 29.6G -r out.json run protection::scenario=mem-hog,loops=1
  ...
  Memory Hog Summary
  ==================

  IO Latency: R p50=123u:84.4u/985u p90=322u:256u/2.5m p99=1.6m:1.4m/9.5m max=11.1m:36.0m/350m
              W p50=429u:274u/995u p90=1.7m:1.3m/4.5m p99=3.4m:2.7m/11.5m max=7.9m:5.9m/26.5m

  Isolation and Request Latency Impact Distributions:

                min   p01   p05   p10   p25   p50   p75   p90   p95   p99   max  mean stdev
  isol%       84.91 84.91 89.51 90.73 92.31 94.49 96.36 98.04 98.71 100.0 100.0 94.42  2.81
  lat-imp%        0     0     0     0     0  2.81  5.73 11.11 13.92 17.53 22.61  4.10  4.68

  Result: isol=94.42:2.81% lat_imp=4.10%:4.68 work_csv=58.34% missing=0%

Signed-off-by: Tejun Heo <tj@kernel.org>
Fixes: a647a524a467 ("block: don't call rq_qos_ops->done_bio if the bio isn't tracked")
Cc: stable@vger.kernel.org # v5.15+
Cc: Ming Lei <ming.lei@redhat.com>
Cc: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Link: https://lore.kernel.org/r/Yi7rdrzQEHjJLGKB@slm.duckdns.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoblock: only mark bio as tracked if it really is tracked
Jens Axboe [Sat, 16 Oct 2021 02:06:18 +0000 (20:06 -0600)]
block: only mark bio as tracked if it really is tracked

[ Upstream commit 90b8faa0e8de1b02b619fb33f6c6e1e13e7d1d70 ]

We set BIO_TRACKED unconditionally when rq_qos_throttle() is called, even
though we may not even have an rq_qos handler. Only mark it as TRACKED if
it really is potentially tracked.

This saves considerable time for the case where the bio isn't tracked:

     2.64%     -1.65%  [kernel.vmlinux]  [k] bio_endio

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoblock: use bdev_get_queue() in bio.c
Pavel Begunkov [Thu, 14 Oct 2021 14:03:28 +0000 (15:03 +0100)]
block: use bdev_get_queue() in bio.c

[ Upstream commit 3caee4634be68e755d2fb130962f1623661dbd5b ]

Convert bdev->bd_disk->queue to bdev_get_queue(), it's uses a cached
queue pointer and so is faster.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/85c36ea784d285a5075baa10049e6b59e15fb484.1634219547.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoio_uring: ensure that fsnotify is always called
Jens Axboe [Sun, 20 Mar 2022 19:08:38 +0000 (13:08 -0600)]
io_uring: ensure that fsnotify is always called

[ Upstream commit f63cf5192fe3418ad5ae1a4412eba5694b145f79 ]

Ensure that we call fsnotify_modify() if we write a file, and that we
do fsnotify_access() if we read it. This enables anyone using inotify
on the file to get notified.

Ditto for fallocate, ensure that fsnotify_modify() is called.

Cc: stable@vger.kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agovirtio-blk: avoid preallocating big SGL for data
Max Gurtovoy [Wed, 1 Sep 2021 13:14:34 +0000 (16:14 +0300)]
virtio-blk: avoid preallocating big SGL for data

[ Upstream commit 02746e26c39ee473b975e0f68d1295abc92672ed ]

No need to pre-allocate a big buffer for the IO SGL anymore. If a device
has lots of deep queues, preallocation for the sg list can consume
substantial amounts of memory. For HW virtio-blk device, nr_hw_queues
can be 64 or 128 and each queue's depth might be 128. This means the
resulting preallocation for the data SGLs is big.

Switch to runtime allocation for SGL for lists longer than 2 entries.
This is the approach used by NVMe drivers so it should be reasonable for
virtio block as well. Runtime SGL allocation has always been the case
for the legacy I/O path so this is nothing new.

The preallocated small SGL depends on SG_CHAIN so if the ARCH doesn't
support SG_CHAIN, use only runtime allocation for the SGL.

Re-organize the setup of the IO request to fit the new sg chain
mechanism.

No performance degradation was seen (fio libaio engine with 16 jobs and
128 iodepth):

IO size      IOPs Rand Read (before/after)         IOPs Rand Write (before/after)
--------     ---------------------------------    ----------------------------------
512B          318K/316K                                    329K/325K

4KB           323K/321K                                    353K/349K

16KB          199K/208K                                    250K/275K

128KB         36K/36.1K                                    39.2K/41.7K

Signed-off-by: Max Gurtovoy <mgurtovoy@nvidia.com>
Reviewed-by: Israel Rukshin <israelr@nvidia.com>
Link: https://lore.kernel.org/r/20210901131434.31158-1-mgurtovoy@nvidia.com
Reviewed-by: Feng Li <lifeng1519@gmail.com>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Tested-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Arnd Bergmann <arnd@arndb.de> # kconfig fixups
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoibmvnic: Allow queueing resets during probe
Sukadev Bhattiprolu [Fri, 25 Feb 2022 06:23:58 +0000 (22:23 -0800)]
ibmvnic: Allow queueing resets during probe

[ Upstream commit fd98693cb0721317f27341951593712c580c36a1 ]

We currently don't allow queuing resets when adapter is in VNIC_PROBING
state - instead we throw away the reset and return EBUSY. The reasoning
is probably that during ibmvnic_probe() the ibmvnic_adapter itself is
being initialized so performing a reset during this time can lead us to
accessing fields in the ibmvnic_adapter that are not fully initialized.
A review of the code shows that all the adapter state neede to process a
reset is initialized before registering the CRQ so that should no longer
be a concern.

Further the expectation is that if we do get a reset (transport event)
during probe, the do..while() loop in ibmvnic_probe() will handle this
by reinitializing the CRQ.

While that is true to some extent, it is possible that the reset might
occur _after_ the CRQ is registered and CRQ_INIT message was exchanged
but _before_ the adapter state is set to VNIC_PROBED. As mentioned above,
such a reset will be thrown away. While the client assumes that the
adapter is functional, the vnic server will wait for the client to reinit
the adapter. This disconnect between the two leaves the adapter down
needing manual intervention.

Because ibmvnic_probe() has other work to do after initializing the CRQ
(such as registering the netdev at a minimum) and because the reset event
can occur at any instant after the CRQ is initialized, there will always
be a window between initializing the CRQ and considering the adapter
ready for resets (ie state == PROBED).

So rather than discarding resets during this window, allow queueing them
- but only process them after the adapter is fully initialized.

To do this, introduce a new completion state ->probe_done and have the
reset worker thread wait on this before processing resets.

This change brings up two new situations in or just after ibmvnic_probe().
First after one or more resets were queued, we encounter an error and
decide to retry the initialization.  At that point the queued resets are
no longer relevant since we could be talking to a new vnic server. So we
must purge/flush the queued resets before restarting the initialization.
As a side note, since we are still in the probing stage and we have not
registered the netdev, it will not be CHANGE_PARAM reset.

Second this change opens up a potential race between the worker thread
in __ibmvnic_reset(), the tasklet and the ibmvnic_open() due to the
following sequence of events:

1. Register CRQ
2. Get transport event before CRQ_INIT completes.
3. Tasklet schedules reset:
a) add rwi to list
b) schedule_work() to start worker thread which runs
   and waits for ->probe_done.
4. ibmvnic_probe() decides to retry, purges rwi_list
5. Re-register crq and this time rest of probe succeeds - register
   netdev and complete(->probe_done).
6. Worker thread resumes in __ibmvnic_reset() from 3b.
7. Worker thread sets ->resetting bit
8. ibmvnic_open() comes in, notices ->resetting bit, sets state
   to IBMVNIC_OPEN and returns early expecting worker thread to
   finish the open.
9. Worker thread finds rwi_list empty and returns without
   opening the interface.

If this happens, the ->ndo_open() call is effectively lost and the
interface remains down. To address this, ensure that ->rwi_list is
not empty before setting the ->resetting  bit. See also comments in
__ibmvnic_reset().

Fixes: 6a2fb0e99f9c ("ibmvnic: driver initialization for kdump/kexec")
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoibmvnic: clear fop when retrying probe
Sukadev Bhattiprolu [Fri, 25 Feb 2022 06:23:57 +0000 (22:23 -0800)]
ibmvnic: clear fop when retrying probe

[ Upstream commit f628ad531b4f34fdba0984255b4a2850dd369513 ]

Clear ->failover_pending flag that may have been set in the previous
pass of registering CRQ. If we don't clear, a subsequent ibmvnic_open()
call would be misled into thinking a failover is pending and assuming
that the reset worker thread would open the adapter. If this pass of
registering the CRQ succeeds (i.e there is no transport event), there
wouldn't be a reset worker thread.

This would leave the adapter unconfigured and require manual intervention
to bring it up during boot.

Fixes: 5a18e1e0c193 ("ibmvnic: Fix failover case for non-redundant configuration")
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoibmvnic: init init_done_rc earlier
Sukadev Bhattiprolu [Fri, 25 Feb 2022 06:23:56 +0000 (22:23 -0800)]
ibmvnic: init init_done_rc earlier

[ Upstream commit ae16bf15374d8b055e040ac6f3f1147ab1c9bb7d ]

We currently initialize the ->init_done completion/return code fields
before issuing a CRQ_INIT command. But if we get a transport event soon
after registering the CRQ the taskslet may already have recorded the
completion and error code. If we initialize here, we might overwrite/
lose that and end up issuing the CRQ_INIT only to timeout later.

If that timeout happens during probe, we will leave the adapter in the
DOWN state rather than retrying to register/init the CRQ.

Initialize the completion before registering the CRQ so we don't lose
the notification.

Fixes: 032c5e82847a ("Driver for IBM System i/p VNIC protocol")
Signed-off-by: Sukadev Bhattiprolu <sukadev@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agos390/setup: preserve memory at OLDMEM_BASE and OLDMEM_SIZE
Alexander Egorenkov [Wed, 9 Feb 2022 10:25:09 +0000 (11:25 +0100)]
s390/setup: preserve memory at OLDMEM_BASE and OLDMEM_SIZE

[ Upstream commit 6b4b54c7ca347bcb4aa7a3cc01aa16e84ac7fbe4 ]

We need to preserve the values at OLDMEM_BASE and OLDMEM_SIZE which are
used by zgetdump in case when kdump crashes. In that case zgetdump will
attempt to read OLDMEM_BASE and OLDMEM_SIZE in order to find out where
the memory range [0 - OLDMEM_SIZE] belonging to the production kernel is.

Fixes: f1a546947431 ("s390/setup: don't reserve memory that occupied decompressor's head")
Cc: stable@vger.kernel.org # 5.15+
Signed-off-by: Alexander Egorenkov <egorenar@linux.ibm.com>
Acked-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agos390/setup: use physical pointers for memblock_reserve()
Alexander Gordeev [Thu, 21 Jan 2021 12:06:02 +0000 (13:06 +0100)]
s390/setup: use physical pointers for memblock_reserve()

[ Upstream commit 04f11ed7d8e018e1f01ebda5814ddfeb3a1e6ae1 ]

memblock_reserve() function accepts physcal address of a memory
block to be reserved, but provided with virtual memory pointers.

Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agos390/boot: allocate amode31 section in decompressor
Alexander Gordeev [Mon, 27 Sep 2021 12:18:26 +0000 (14:18 +0200)]
s390/boot: allocate amode31 section in decompressor

[ Upstream commit e3ec8e0f5711d73f7e5d5c3cffdf4fad4f1487b9 ]

The memory for amode31 section is allocated from the decompressed
kernel. Instead, allocate that memory from the decompressor. This
is a prerequisite to allow initialization of the virtual memory
before the decompressed kernel takes over.

Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agonetfilter: nft_payload: don't allow th access for fragments
Florian Westphal [Sat, 29 Jan 2022 16:13:23 +0000 (17:13 +0100)]
netfilter: nft_payload: don't allow th access for fragments

[ Upstream commit a9e8503def0fd4ed89ade1f61c315f904581d439 ]

Loads relative to ->thoff naturally expect that this points to the
transport header, but this is only true if pkt->fragoff == 0.

This has little effect for rulesets with connection tracking/nat because
these enable ip defra. For other rulesets this prevents false matches.

Fixes: 96518518cc41 ("netfilter: add nftables")
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agonetfilter: nft_payload: support for inner header matching / mangling
Pablo Neira Ayuso [Thu, 28 Oct 2021 20:15:00 +0000 (22:15 +0200)]
netfilter: nft_payload: support for inner header matching / mangling

[ Upstream commit c46b38dc8743535e686b911d253a844f0bd50ead ]

Allow to match and mangle on inner headers / payload data after the
transport header. There is a new field in the pktinfo structure that
stores the inner header offset which is calculated only when requested.
Only TCP and UDP supported at this stage.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agonetfilter: nf_tables: convert pktinfo->tprot_set to flags field
Pablo Neira Ayuso [Thu, 28 Oct 2021 19:47:55 +0000 (21:47 +0200)]
netfilter: nf_tables: convert pktinfo->tprot_set to flags field

[ Upstream commit b5bdc6f9c24db9a0adf8bd00c0e935b184654f00 ]

Generalize boolean field to store more flags on the pktinfo structure.

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoASoC: rt5682: Fix deadlock on resume
Peter Ujfalusi [Wed, 26 Jan 2022 10:03:25 +0000 (12:03 +0200)]
ASoC: rt5682: Fix deadlock on resume

[ Upstream commit 4045daf0fa87846a27f56329fddad2deeb5ca354 ]

On resume from suspend the following chain of events can happen:
A rt5682_resume() -> mod_delayed_work() for jack_detect_work
B DAPM sequence starts ( DAPM is locked now)

A1. rt5682_jack_detect_handler() scheduled
 - Takes both jdet_mutex and calibrate_mutex
 - Calls in to rt5682_headset_detect() which tries to take DAPM lock, it
   starts to wait for it as B path took it already.
B1. DAPM sequence reaches the "HP Amp", rt5682_hp_event() tries to take
    the jdet_mutex, but it is locked in A1, so it waits.

Deadlock.

To solve the deadlock, drop the jdet_mutex, use the jack_detect_work to do
the jack removal handling, move the dapm lock up one level to protect the
most of the rt5682_jack_detect_handler(), but not the jack reporting as it
might trigger a DAPM sequence.
The rt5682_headset_detect() can be changed to static as well.

Fixes: 8deb34a90f063 ("ASoC: rt5682: fix the wrong jack type detected")
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@linux.intel.com>
Link: https://lore.kernel.org/r/20220126100325.16513-1-peter.ujfalusi@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoASoC: rt5682: Re-detect the combo jack after resuming
Derek Fang [Tue, 9 Nov 2021 09:54:50 +0000 (17:54 +0800)]
ASoC: rt5682: Re-detect the combo jack after resuming

[ Upstream commit 2cd9b0ef82d936623d789bb3fbb6fcf52c500367 ]

Sometimes, end-users change the jack type under suspending,
so it needs to re-detect the combo jack type after resuming to
avoid any unexpected behaviors.

Signed-off-by: Derek Fang <derek.fang@realtek.com>
Link: https://lore.kernel.org/r/20211109095450.12950-2-derek.fang@realtek.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoASoC: rt5682: Avoid the unexpected IRQ event during going to suspend
Derek Fang [Tue, 9 Nov 2021 09:54:49 +0000 (17:54 +0800)]
ASoC: rt5682: Avoid the unexpected IRQ event during going to suspend

[ Upstream commit a3774a2a6544a7a4a85186e768afc07044aa507f ]

When the system suspends, the codec driver will set SAR to
power saving mode if a headset is plugged in.
There is a chance to generate an unexpected IRQ, and leads to
issues after resuming such as noise from OMTP type headsets.

Signed-off-by: Derek Fang <derek.fang@realtek.com>
Link: https://lore.kernel.org/r/20211109095450.12950-1-derek.fang@realtek.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agonet/mlx5e: TC, Reject rules with forward and drop actions
Roi Dayan [Mon, 17 Jan 2022 13:00:30 +0000 (15:00 +0200)]
net/mlx5e: TC, Reject rules with forward and drop actions

[ Upstream commit 5623ef8a118838aae65363750dfafcba734dc8cb ]

Such rules are redundant but allowed and passed to the driver.
The driver does not support offloading such rules so return an error.

Fixes: 03a9d11e6eeb ("net/mlx5e: Add TC drop and mirred/redirect action parsing for SRIOV offloads")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Oz Shlomo <ozsh@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agonet/mlx5e: TC, Reject rules with drop and modify hdr action
Roi Dayan [Tue, 4 Jan 2022 08:38:02 +0000 (10:38 +0200)]
net/mlx5e: TC, Reject rules with drop and modify hdr action

[ Upstream commit a2446bc77a16cefd27de712d28af2396d6287593 ]

This kind of action is not supported by firmware and generates a
syndrome.

kernel: mlx5_core 0000:08:00.0: mlx5_cmd_check:777:(pid 102063): SET_FLOW_TABLE_ENTRY(0x936) op_mod(0x0) failed, status bad parameter(0x3), syndrome (0x8708c3)

Fixes: d7e75a325cb2 ("net/mlx5e: Add offloading of E-Switch TC pedit (header re-write) actions")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Oz Shlomo <ozsh@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agonet/mlx5e: Split actions_match_supported() into a sub function
Roi Dayan [Sun, 15 Aug 2021 09:53:13 +0000 (12:53 +0300)]
net/mlx5e: Split actions_match_supported() into a sub function

[ Upstream commit 9c1d3511a2c2fd30c991a20c670991ece4ef27c1 ]

There will probably be more checks, some for nic flows, some for fdb
flows and some are shared checks. Split it for fdb and nic to avoid
the function getting too big.

Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agonet/mlx5e: Check action fwd/drop flag exists also for nic flows
Roi Dayan [Thu, 12 Aug 2021 06:38:32 +0000 (09:38 +0300)]
net/mlx5e: Check action fwd/drop flag exists also for nic flows

[ Upstream commit 6b50cf45b6a0e99f3cab848a72ecca8da56b7460 ]

The driver should add offloaded rules with either a fwd or drop action.
The check existed in parsing fdb flows but not when parsing nic flows.
Move the test into actions_match_supported() which is called for
checking nic flows and fdb flows.

Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Maor Dickman <maord@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoRISC-V: defconfigs: Set CONFIG_FB=y, for FB console
Palmer Dabbelt [Fri, 19 Nov 2021 16:44:02 +0000 (08:44 -0800)]
RISC-V: defconfigs: Set CONFIG_FB=y, for FB console

[ Upstream commit 3d12b634fe8206ea974c6061a3f3eea529ffbc48 ]

We have CONFIG_FRAMEBUFFER_CONSOLE=y in the defconfigs, but that depends
on CONFIG_FB so it's not actually getting set.  I'm assuming most users
on real systems want a framebuffer console, so this enables CONFIG_FB to
allow that to take effect.

Fixes: 33c57c0d3c67 ("RISC-V: Add a basic defconfig")
Reviewed-by: Anup Patel <anup@brainfault.org>
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoriscv: defconfig: enable DRM_NOUVEAU
Heinrich Schuchardt [Tue, 12 Oct 2021 16:46:58 +0000 (18:46 +0200)]
riscv: defconfig: enable DRM_NOUVEAU

[ Upstream commit ffa7a9141bb70702744a312f904b190ca064bdd7 ]

Both RADEON and NOUVEAU graphics cards are supported on RISC-V. Enabling
the one and not the other does not make sense.

As typically at most one of RADEON, NOUVEAU, or VIRTIO GPU support will be
needed DRM drivers should be compiled as modules.

Signed-off-by: Heinrich Schuchardt <heinrich.schuchardt@canonical.com>
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobpf, arm64: Use emit_addr_mov_i64() for BPF_PSEUDO_FUNC
Hou Tao [Fri, 31 Dec 2021 15:10:18 +0000 (23:10 +0800)]
bpf, arm64: Use emit_addr_mov_i64() for BPF_PSEUDO_FUNC

[ Upstream commit e4a41c2c1fa916547e63440c73a51a5eb06247af ]

The following error is reported when running "./test_progs -t for_each"
under arm64:

  bpf_jit: multi-func JIT bug 58 != 56
  [...]
  JIT doesn't support bpf-to-bpf calls

The root cause is the size of BPF_PSEUDO_FUNC instruction increases
from 2 to 3 after the address of called bpf-function is settled and
there are two bpf-to-bpf calls in test_pkt_access. The generated
instructions are shown below:

  0x48:  21 00 C0 D2    movz x1, #0x1, lsl #32
  0x4c:  21 00 80 F2    movk x1, #0x1

  0x48:  E1 3F C0 92    movn x1, #0x1ff, lsl #32
  0x4c:  41 FE A2 F2    movk x1, #0x17f2, lsl #16
  0x50:  81 70 9F F2    movk x1, #0xfb84

Fixing it by using emit_addr_mov_i64() for BPF_PSEUDO_FUNC, so
the size of jited image will not change.

Fixes: 69c087ba6225 ("bpf: Add bpf_for_each_map_elem() helper")
Signed-off-by: Hou Tao <houtao1@huawei.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Link: https://lore.kernel.org/bpf/20211231151018.3781550-1-houtao1@huawei.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobpf: Stop caching subprog index in the bpf_pseudo_func insn
Martin KaFai Lau [Sat, 6 Nov 2021 01:40:14 +0000 (18:40 -0700)]
bpf: Stop caching subprog index in the bpf_pseudo_func insn

commit 3990ed4c426652fcd469f8c9dc08156294b36c28 upstream.

This patch is to fix an out-of-bound access issue when jit-ing the
bpf_pseudo_func insn (i.e. ld_imm64 with src_reg == BPF_PSEUDO_FUNC)

In jit_subprog(), it currently reuses the subprog index cached in
insn[1].imm.  This subprog index is an index into a few array related
to subprogs.  For example, in jit_subprog(), it is an index to the newly
allocated 'struct bpf_prog **func' array.

The subprog index was cached in insn[1].imm after add_subprog().  However,
this could become outdated (and too big in this case) if some subprogs
are completely removed during dead code elimination (in
adjust_subprog_starts_after_remove).  The cached index in insn[1].imm
is not updated accordingly and causing out-of-bound issue in the later
jit_subprog().

Unlike bpf_pseudo_'func' insn, the current bpf_pseudo_'call' insn
is handling the DCE properly by calling find_subprog(insn->imm) to
figure out the index instead of caching the subprog index.
The existing bpf_adj_branches() will adjust the insn->imm
whenever insn is added or removed.

Instead of having two ways handling subprog index,
this patch is to make bpf_pseudo_func works more like
bpf_pseudo_call.

First change is to stop caching the subprog index result
in insn[1].imm after add_subprog().  The verification
process will use find_subprog(insn->imm) to figure
out the subprog index.

Second change is in bpf_adj_branches() and have it to
adjust the insn->imm for the bpf_pseudo_func insn also
whenever insn is added or removed.

Third change is in jit_subprog().  Like the bpf_pseudo_call handling,
bpf_pseudo_func temporarily stores the find_subprog() result
in insn->off.  It is fine because the prog's insn has been finalized
at this point.  insn->off will be reset back to 0 later to avoid
confusing the userspace prog dump tool.

Fixes: 69c087ba6225 ("bpf: Add bpf_for_each_map_elem() helper")
Signed-off-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
Link: https://lore.kernel.org/bpf/20211106014014.651018-1-kafai@fb.com
Cc: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agomt76: mt7921: fix a possible race enabling/disabling runtime-pm
Lorenzo Bianconi [Tue, 21 Dec 2021 13:57:09 +0000 (14:57 +0100)]
mt76: mt7921: fix a possible race enabling/disabling runtime-pm

[ Upstream commit d430dffbe9dd30759f3c64b65bf85b0245c8d8ab ]

Fix a possible race enabling/disabling runtime-pm between
mt7921_pm_set() and mt7921_poll_rx() since mt7921_pm_wake_work()
always schedules rx-napi callback and it will trigger
mt7921_pm_power_save_work routine putting chip to in low-power state
during mt7921_pm_set processing.

Suggested-by: Deren Wu <deren.wu@mediatek.com>
Tested-by: Deren Wu <deren.wu@mediatek.com>
Fixes: 1d8efc741df8 ("mt76: mt7921: introduce Runtime PM support")
Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Kalle Valo <kvalo@kernel.org>
Link: https://lore.kernel.org/r/0f3e075a2033dc05f09dab4059e5be8cbdccc239.1640094847.git.lorenzo@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomt76: mt7921: introduce mt7921_mcu_set_beacon_filter utility routine
Lorenzo Bianconi [Mon, 9 Aug 2021 10:37:22 +0000 (12:37 +0200)]
mt76: mt7921: introduce mt7921_mcu_set_beacon_filter utility routine

[ Upstream commit 890809ca1986e63d29dd1591090af67b655ed89c ]

Introduce mt7921_mcu_set_beacon_filter utility routine in order to
remove duplicated code for hw beacon filtering.
Move mt7921_pm_interface_iter in debugfs since it is just used there.
Make the following routine static:
- mt7921_pm_interface_iter
- mt7921_mcu_uni_bss_bcnft
- mt7921_mcu_set_bss_pm

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomt76: mt7921: get rid of mt7921_mac_set_beacon_filter
Lorenzo Bianconi [Mon, 9 Aug 2021 08:38:03 +0000 (10:38 +0200)]
mt76: mt7921: get rid of mt7921_mac_set_beacon_filter

[ Upstream commit b30363102a4122f6eed37927b64a2c7ac70b8859 ]

Remove mt7921_mac_set_beacon_filter routine since it is no longer used.

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
Signed-off-by: Felix Fietkau <nbd@nbd.name>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoplatform/x86: wmi: Fix driver->notify() vs ->probe() race
Hans de Goede [Sun, 28 Nov 2021 19:00:28 +0000 (20:00 +0100)]
platform/x86: wmi: Fix driver->notify() vs ->probe() race

[ Upstream commit 9918878676a5f9e99b98679f04b9e6c0f5426b0a ]

The driver core sets struct device->driver before calling out
to the bus' probe() method, this leaves a window where an ACPI
notify may happen on the WMI object before the driver's
probe() method has completed running, causing e.g. the
driver's notify() callback to get called with drvdata
not yet being set leading to a NULL pointer deref.

At a check for this to the WMI core, ensuring that the notify()
callback is not called before the driver is ready.

Fixes: 1686f5444546 ("platform/x86: wmi: Incorporate acpi_install_notify_handler")
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20211128190031.405620-2-hdegoede@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoplatform/x86: wmi: Replace read_takes_no_args with a flags field
Hans de Goede [Sun, 28 Nov 2021 19:00:27 +0000 (20:00 +0100)]
platform/x86: wmi: Replace read_takes_no_args with a flags field

[ Upstream commit a90b38c58667142ecff2521481ed44286d46b140 ]

Replace the wmi_block.read_takes_no_args bool field with
an unsigned long flags field, used together with test_bit()
and friends.

This is a preparation patch for fixing a driver->notify() vs ->probe()
race, which requires atomic flag handling.

Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20211128190031.405620-1-hdegoede@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoplatform/x86: wmi: introduce helper to convert driver to WMI driver
Barnabás Pőcze [Sat, 4 Sep 2021 17:56:32 +0000 (17:56 +0000)]
platform/x86: wmi: introduce helper to convert driver to WMI driver

[ Upstream commit e7b2e33449e22fdbaa0247d96f31543affe6163d ]

Introduce a helper function which wraps the appropriate
`container_of()` macro invocation to convert
a `struct device_driver` to `struct wmi_driver`.

Signed-off-by: Barnabás Pőcze <pobrn@protonmail.com>
Link: https://lore.kernel.org/r/20210904175450.156801-27-pobrn@protonmail.com
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoqed: Improve the stack space of filter_config()
Shai Malin [Mon, 13 Sep 2021 07:50:24 +0000 (10:50 +0300)]
qed: Improve the stack space of filter_config()

[ Upstream commit f55e36d5ab76c3097ff36ecea60b91c6b0d80fc8 ]

As it was reported and discussed in: https://lore.kernel.org/lkml/CAHk-=whF9F89vsfH8E9TGc0tZA-yhzi2Di8wOtquNB5vRkFX5w@mail.gmail.com/
This patch improves the stack space of qede_config_rx_mode() by
splitting filter_config() to 3 functions and removing the
union qed_filter_type_params.

Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Signed-off-by: Ariel Elior <aelior@marvell.com>
Signed-off-by: Shai Malin <smalin@marvell.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoath11k: add hw_param for wakeup_mhi
Seevalamuthu Mariappan [Wed, 17 Nov 2021 07:39:41 +0000 (09:39 +0200)]
ath11k: add hw_param for wakeup_mhi

[ Upstream commit 081e2d6476e30399433b509684d5da4d1844e430 ]

Wakeup mhi is needed before pci_read/write only for QCA6390 and WCN6855. Since
wakeup & release mhi is enabled for all hardwares, below mhi assert is seen in
QCN9074 when doing 'rmmod ath11k_pci':

Kernel panic - not syncing: dev_wake != 0
CPU: 2 PID: 13535 Comm: procd Not tainted 4.4.60 #1
Hardware name: Generic DT based system
[<80316dac>] (unwind_backtrace) from [<80313700>] (show_stack+0x10/0x14)
[<80313700>] (show_stack) from [<805135dc>] (dump_stack+0x7c/0x9c)
[<805135dc>] (dump_stack) from [<8032136c>] (panic+0x84/0x1f8)
[<8032136c>] (panic) from [<80549b24>] (mhi_pm_disable_transition+0x3b8/0x5b8)
[<80549b24>] (mhi_pm_disable_transition) from [<80549ddc>] (mhi_power_down+0xb8/0x100)
[<80549ddc>] (mhi_power_down) from [<7f5242b0>] (ath11k_mhi_op_status_cb+0x284/0x3ac [ath11k_pci])
[E][__mhi_device_get_sync] Did not enter M0 state, cur_state:RESET pm_state:SHUTDOWN Process
[E][__mhi_device_get_sync] Did not enter M0 state, cur_state:RESET pm_state:SHUTDOWN Process
[E][__mhi_device_get_sync] Did not enter M0 state, cur_state:RESET pm_state:SHUTDOWN Process
[<7f5242b0>] (ath11k_mhi_op_status_cb [ath11k_pci]) from [<7f524878>] (ath11k_mhi_stop+0x10/0x20 [ath11k_pci])
[<7f524878>] (ath11k_mhi_stop [ath11k_pci]) from [<7f525b94>] (ath11k_pci_power_down+0x54/0x90 [ath11k_pci])
[<7f525b94>] (ath11k_pci_power_down [ath11k_pci]) from [<8056b2a8>] (pci_device_shutdown+0x30/0x44)
[<8056b2a8>] (pci_device_shutdown) from [<805cfa0c>] (device_shutdown+0x124/0x174)
[<805cfa0c>] (device_shutdown) from [<8033aaa4>] (kernel_restart+0xc/0x50)
[<8033aaa4>] (kernel_restart) from [<8033ada8>] (SyS_reboot+0x178/0x1ec)
[<8033ada8>] (SyS_reboot) from [<80301b80>] (ret_fast_syscall+0x0/0x34)

Hence, disable wakeup/release mhi using hw_param for other hardwares.

Tested-on: QCN9074 hw1.0 PCI WLAN.HK.2.5.0.1-01060-QCAHKSWPL_SILICONZ-1

Fixes: a05bd8513335 ("ath11k: read and write registers below unwindowed address")
Signed-off-by: Seevalamuthu Mariappan <quic_seevalam@quicinc.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/1636702019-26142-1-git-send-email-quic_seevalam@quicinc.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomemory: renesas-rpc-if: Avoid unaligned bus access for HyperFlash
Andrew Gabbasov [Wed, 22 Sep 2021 18:48:30 +0000 (13:48 -0500)]
memory: renesas-rpc-if: Avoid unaligned bus access for HyperFlash

[ Upstream commit 1869023e24c0de73a160a424dac4621cefd628ae ]

HyperFlash devices in Renesas SoCs use 2-bytes addressing, according
to HW manual paragraph 62.3.3 (which officially describes Serial Flash
access, but seems to be applicable to HyperFlash too). And 1-byte bus
read operations to 2-bytes unaligned addresses in external address space
read mode work incorrectly (returns the other byte from the same word).

Function memcpy_fromio(), used by the driver to read data from the bus,
in ARM64 architecture (to which Renesas cores belong) uses 8-bytes
bus accesses for appropriate aligned addresses, and 1-bytes accesses
for other addresses. This results in incorrect data read from HyperFlash
in unaligned cases.

This issue can be reproduced using something like the following commands
(where mtd1 is a parition on Hyperflash storage, defined properly
in a device tree):

[Correct fragment, read from Hyperflash]

    root@rcar-gen3:~# dd if=/dev/mtd1 of=/tmp/zz bs=32 count=1
    root@rcar-gen3:~# hexdump -C /tmp/zz
    00000000  f4 03 00 aa f5 03 01 aa  f6 03 02 aa f7 03 03 aa  |................|
    00000010  00 00 80 d2 40 20 18 d5  00 06 81 d2 a0 18 a6 f2  |....@ ..........|
    00000020

[Incorrect read of the same fragment: see the difference at offsets 8-11]

    root@rcar-gen3:~# dd if=/dev/mtd1 of=/tmp/zz bs=12 count=1
    root@rcar-gen3:~# hexdump -C /tmp/zz
    00000000  f4 03 00 aa f5 03 01 aa  03 03 aa aa              |............|
    0000000c

Fix this issue by creating a local replacement of the copying function,
that performs only properly aligned bus accesses, and is used for reading
from HyperFlash.

Fixes: ca7d8b980b67f ("memory: add Renesas RPC-IF driver")
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Gabbasov <andrew_gabbasov@mentor.com>
Link: https://lore.kernel.org/r/20210922184830.29147-1-andrew_gabbasov@mentor.com
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@canonical.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agomedia: ir_toy: prevent device from hanging during transmit
Sean Young [Tue, 14 Sep 2021 14:57:46 +0000 (16:57 +0200)]
media: ir_toy: prevent device from hanging during transmit

[ Upstream commit 4114978dcd24e72415276bba60ff4ff355970bbc ]

If the IR Toy is receiving IR while a transmit is done, it may end up
hanging. We can prevent this from happening by re-entering sample mode
just before issuing the transmit command.

Link: https://github.com/bengtmartensson/HarcHardware/discussions/25
Cc: stable@vger.kernel.org
[mchehab: renamed: s/STATE_RESET/STATE_COMMAND_NO_RESP/ ]
Signed-off-by: Sean Young <sean@mess.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoPCI: pciehp: Ignore Link Down/Up caused by error-induced Hot Reset
Lukas Wunner [Sat, 31 Jul 2021 12:39:01 +0000 (14:39 +0200)]
PCI: pciehp: Ignore Link Down/Up caused by error-induced Hot Reset

[ Upstream commit ea401499e943c307e6d44af6c2b4e068643e7884 ]

Stuart Hayes reports that an error handled by DPC at a Root Port results
in pciehp gratuitously bringing down a subordinate hotplug port:

  RP -- UP -- DP -- UP -- DP (hotplug) -- EP

pciehp brings the slot down because the Link to the Endpoint goes down.
That is caused by a Hot Reset being propagated as a result of DPC.
Per PCIe Base Spec 5.0, section 6.6.1 "Conventional Reset":

  For a Switch, the following must cause a hot reset to be sent on all
  Downstream Ports: [...]

  * The Data Link Layer of the Upstream Port reporting DL_Down status.
    In Switches that support Link speeds greater than 5.0 GT/s, the
    Upstream Port must direct the LTSSM of each Downstream Port to the
    Hot Reset state, but not hold the LTSSMs in that state. This permits
    each Downstream Port to begin Link training immediately after its
    hot reset completes. This behavior is recommended for all Switches.

  * Receiving a hot reset on the Upstream Port.

Once DPC recovers, pcie_do_recovery() walks down the hierarchy and
invokes pcie_portdrv_slot_reset() to restore each port's config space.
At that point, a hotplug interrupt is signaled per PCIe Base Spec r5.0,
section 6.7.3.4 "Software Notification of Hot-Plug Events":

  If the Port is enabled for edge-triggered interrupt signaling using
  MSI or MSI-X, an interrupt message must be sent every time the logical
  AND of the following conditions transitions from FALSE to TRUE: [...]

  * The Hot-Plug Interrupt Enable bit in the Slot Control register is
    set to 1b.

  * At least one hot-plug event status bit in the Slot Status register
    and its associated enable bit in the Slot Control register are both
    set to 1b.

Prevent pciehp from gratuitously bringing down the slot by clearing the
error-induced Data Link Layer State Changed event before restoring
config space.  Afterwards, check whether the link has unexpectedly
failed to retrain and synthesize a DLLSC event if so.

Allow each pcie_port_service_driver (one of them being pciehp) to define
a slot_reset callback and re-use the existing pm_iter() function to
iterate over the callbacks.

Thereby, the Endpoint driver remains bound throughout error recovery and
may restore the device to working state.

Surprise removal during error recovery is detected through a Presence
Detect Changed event.  The hotplug port is expected to not signal that
event as a result of a Hot Reset.

The issue isn't DPC-specific, it also occurs when an error is handled by
AER through aer_root_reset().  So while the issue was noticed only now,
it's been around since 2006 when AER support was first introduced.

[bhelgaas: drop PCI_ERROR_RECOVERY Kconfig, split pm_iter() rename to
preparatory patch]
Link: https://lore.kernel.org/linux-pci/08c046b0-c9f2-3489-eeef-7e7aca435bb9@gmail.com/
Fixes: 6c2b374d7485 ("PCI-Express AER implemetation: AER core and aerdriver")
Link: https://lore.kernel.org/r/251f4edcc04c14f873ff1c967bc686169cd07d2d.1627638184.git.lukas@wunner.de
Reported-by: Stuart Hayes <stuart.w.hayes@gmail.com>
Tested-by: Stuart Hayes <stuart.w.hayes@gmail.com>
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org # v2.6.19+: ba952824e6c1: PCI/portdrv: Report reset for frozen channel
Cc: Keith Busch <kbusch@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoPCI/portdrv: Rename pm_iter() to pcie_port_device_iter()
Lukas Wunner [Fri, 15 Oct 2021 18:58:40 +0000 (13:58 -0500)]
PCI/portdrv: Rename pm_iter() to pcie_port_device_iter()

[ Upstream commit 3134689f98f9e09004a4727370adc46e7635b4be ]

Rename pm_iter() to pcie_port_device_iter() and make it visible outside
CONFIG_PM and portdrv_core.c so it can be used for pciehp slot reset
recovery.

[bhelgaas: split into its own patch]
Link: https://lore.kernel.org/linux-pci/08c046b0-c9f2-3489-eeef-7e7aca435bb9@gmail.com/
Link: https://lore.kernel.org/r/251f4edcc04c14f873ff1c967bc686169cd07d2d.1627638184.git.lukas@wunner.de
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agodrm/i915: Replace the unconditional clflush with drm_clflush_virt_range()
Ville Syrjälä [Thu, 14 Oct 2021 09:09:38 +0000 (12:09 +0300)]
drm/i915: Replace the unconditional clflush with drm_clflush_virt_range()

[ Upstream commit ef7ec41f17cbc0861891ccc0634d06a0c8dcbf09 ]

Not all machines have clflush, so don't go assuming they do.
Not really sure why the clflush is even here since hwsp
is supposed to get snooped I thought.

Although in my case we're talking about a i830 machine where
render/blitter snooping is definitely busted. But it might
work for the hswp perhaps. Haven't really reverse engineered
that one fully.

Cc: stable@vger.kernel.org
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Fixes: b436a5f8b6c8 ("drm/i915/gt: Track all timelines created using the HWSP")
Signed-off-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211014090941.12159-2-ville.syrjala@linux.intel.com
Reviewed-by: Dave Airlie <airlied@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agodrm/i915/gt: Register the migrate contexts with their engines
Thomas Hellström [Wed, 22 Sep 2021 06:25:23 +0000 (08:25 +0200)]
drm/i915/gt: Register the migrate contexts with their engines

[ Upstream commit 3e42cc61275f95fd7f022b6380b95428efe134d3 ]

Pinned contexts, like the migrate contexts need reset after resume
since their context image may have been lost. Also the GuC needs to
register pinned contexts.

Add a list to struct intel_engine_cs where we add all pinned contexts on
creation, and traverse that list at resume time to reset the pinned
contexts.

This fixes the kms_pipe_crc_basic@suspend-read-crc-pipe-a selftest for now,
but proper LMEM backup / restore is needed for full suspend functionality.
However, note that even with full LMEM backup / restore it may be
desirable to keep the reset since backing up the migrate context images
must happen using memcpy() after the migrate context has become inactive,
and for performance- and other reasons we want to avoid memcpy() from
LMEM.

Also traverse the list at guc_init_lrc_mapping() calling
guc_kernel_context_pin() for the pinned contexts, like is already done
for the kernel context.

v2:
- Don't reset the contexts on each __engine_unpark() but rather at
  resume time (Chris Wilson).
v3:
- Reset contexts in the engine sanitize callback. (Chris Wilson)

Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Matthew Auld <matthew.auld@intel.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Brost Matthew <matthew.brost@intel.com>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Signed-off-by: Thomas Hellström <thomas.hellstrom@linux.intel.com>
Reviewed-by: Matthew Auld <matthew.auld@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210922062527.865433-6-thomas.hellstrom@linux.intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agodrm/i915: Disable bonding on gen12+ platforms
Matthew Brost [Wed, 28 Jul 2021 19:21:00 +0000 (12:21 -0700)]
drm/i915: Disable bonding on gen12+ platforms

[ Upstream commit ce7e75c7ef1bf8ea3d947da8c674d2f40fd7d734 ]

Disable bonding on gen12+ platforms aside from ones already supported by
the i915 - TGL, RKL, and ADL-S.

Signed-off-by: Matthew Brost <matthew.brost@intel.com>
Reviewed-by: John Harrison <John.C.Harrison@Intel.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210728192100.132425-1-matthew.brost@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: fix deadlock between chunk allocation and chunk btree modifications
Filipe Manana [Wed, 13 Oct 2021 09:12:49 +0000 (10:12 +0100)]
btrfs: fix deadlock between chunk allocation and chunk btree modifications

[ Upstream commit 2bb2e00ed9787e52580bb651264b8d6a2b7a9dd2 ]

When a task is doing some modification to the chunk btree and it is not in
the context of a chunk allocation or a chunk removal, it can deadlock with
another task that is currently allocating a new data or metadata chunk.

These contexts are the following:

* When relocating a system chunk, when we need to COW the extent buffers
  that belong to the chunk btree;

* When adding a new device (ioctl), where we need to add a new device item
  to the chunk btree;

* When removing a device (ioctl), where we need to remove a device item
  from the chunk btree;

* When resizing a device (ioctl), where we need to update a device item in
  the chunk btree and may need to relocate a system chunk that lies beyond
  the new device size when shrinking a device.

The problem happens due to a sequence of steps like the following:

1) Task A starts a data or metadata chunk allocation and it locks the
   chunk mutex;

2) Task B is relocating a system chunk, and when it needs to COW an extent
   buffer of the chunk btree, it has locked both that extent buffer as
   well as its parent extent buffer;

3) Since there is not enough available system space, either because none
   of the existing system block groups have enough free space or because
   the only one with enough free space is in RO mode due to the relocation,
   task B triggers a new system chunk allocation. It blocks when trying to
   acquire the chunk mutex, currently held by task A;

4) Task A enters btrfs_chunk_alloc_add_chunk_item(), in order to insert
   the new chunk item into the chunk btree and update the existing device
   items there. But in order to do that, it has to lock the extent buffer
   that task B locked at step 2, or its parent extent buffer, but task B
   is waiting on the chunk mutex, which is currently locked by task A,
   therefore resulting in a deadlock.

One example report when the deadlock happens with system chunk relocation:

  INFO: task kworker/u9:5:546 blocked for more than 143 seconds.
        Not tainted 5.15.0-rc3+ #1
  "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
  task:kworker/u9:5    state:D stack:25936 pid:  546 ppid:     2 flags:0x00004000
  Workqueue: events_unbound btrfs_async_reclaim_metadata_space
  Call Trace:
   context_switch kernel/sched/core.c:4940 [inline]
   __schedule+0xcd9/0x2530 kernel/sched/core.c:6287
   schedule+0xd3/0x270 kernel/sched/core.c:6366
   rwsem_down_read_slowpath+0x4ee/0x9d0 kernel/locking/rwsem.c:993
   __down_read_common kernel/locking/rwsem.c:1214 [inline]
   __down_read kernel/locking/rwsem.c:1223 [inline]
   down_read_nested+0xe6/0x440 kernel/locking/rwsem.c:1590
   __btrfs_tree_read_lock+0x31/0x350 fs/btrfs/locking.c:47
   btrfs_tree_read_lock fs/btrfs/locking.c:54 [inline]
   btrfs_read_lock_root_node+0x8a/0x320 fs/btrfs/locking.c:191
   btrfs_search_slot_get_root fs/btrfs/ctree.c:1623 [inline]
   btrfs_search_slot+0x13b4/0x2140 fs/btrfs/ctree.c:1728
   btrfs_update_device+0x11f/0x500 fs/btrfs/volumes.c:2794
   btrfs_chunk_alloc_add_chunk_item+0x34d/0xea0 fs/btrfs/volumes.c:5504
   do_chunk_alloc fs/btrfs/block-group.c:3408 [inline]
   btrfs_chunk_alloc+0x84d/0xf50 fs/btrfs/block-group.c:3653
   flush_space+0x54e/0xd80 fs/btrfs/space-info.c:670
   btrfs_async_reclaim_metadata_space+0x396/0xa90 fs/btrfs/space-info.c:953
   process_one_work+0x9df/0x16d0 kernel/workqueue.c:2297
   worker_thread+0x90/0xed0 kernel/workqueue.c:2444
   kthread+0x3e5/0x4d0 kernel/kthread.c:319
   ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:295
  INFO: task syz-executor:9107 blocked for more than 143 seconds.
        Not tainted 5.15.0-rc3+ #1
  "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
  task:syz-executor    state:D stack:23200 pid: 9107 ppid:  7792 flags:0x00004004
  Call Trace:
   context_switch kernel/sched/core.c:4940 [inline]
   __schedule+0xcd9/0x2530 kernel/sched/core.c:6287
   schedule+0xd3/0x270 kernel/sched/core.c:6366
   schedule_preempt_disabled+0xf/0x20 kernel/sched/core.c:6425
   __mutex_lock_common kernel/locking/mutex.c:669 [inline]
   __mutex_lock+0xc96/0x1680 kernel/locking/mutex.c:729
   btrfs_chunk_alloc+0x31a/0xf50 fs/btrfs/block-group.c:3631
   find_free_extent_update_loop fs/btrfs/extent-tree.c:3986 [inline]
   find_free_extent+0x25cb/0x3a30 fs/btrfs/extent-tree.c:4335
   btrfs_reserve_extent+0x1f1/0x500 fs/btrfs/extent-tree.c:4415
   btrfs_alloc_tree_block+0x203/0x1120 fs/btrfs/extent-tree.c:4813
   __btrfs_cow_block+0x412/0x1620 fs/btrfs/ctree.c:415
   btrfs_cow_block+0x2f6/0x8c0 fs/btrfs/ctree.c:570
   btrfs_search_slot+0x1094/0x2140 fs/btrfs/ctree.c:1768
   relocate_tree_block fs/btrfs/relocation.c:2694 [inline]
   relocate_tree_blocks+0xf73/0x1770 fs/btrfs/relocation.c:2757
   relocate_block_group+0x47e/0xc70 fs/btrfs/relocation.c:3673
   btrfs_relocate_block_group+0x48a/0xc60 fs/btrfs/relocation.c:4070
   btrfs_relocate_chunk+0x96/0x280 fs/btrfs/volumes.c:3181
   __btrfs_balance fs/btrfs/volumes.c:3911 [inline]
   btrfs_balance+0x1f03/0x3cd0 fs/btrfs/volumes.c:4301
   btrfs_ioctl_balance+0x61e/0x800 fs/btrfs/ioctl.c:4137
   btrfs_ioctl+0x39ea/0x7b70 fs/btrfs/ioctl.c:4949
   vfs_ioctl fs/ioctl.c:51 [inline]
   __do_sys_ioctl fs/ioctl.c:874 [inline]
   __se_sys_ioctl fs/ioctl.c:860 [inline]
   __x64_sys_ioctl+0x193/0x200 fs/ioctl.c:860
   do_syscall_x64 arch/x86/entry/common.c:50 [inline]
   do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
   entry_SYSCALL_64_after_hwframe+0x44/0xae

So fix this by making sure that whenever we try to modify the chunk btree
and we are neither in a chunk allocation context nor in a chunk remove
context, we reserve system space before modifying the chunk btree.

Reported-by: Hao Sun <sunhao.th@gmail.com>
Link: https://lore.kernel.org/linux-btrfs/CACkBjsax51i4mu6C0C3vJqQN3NR_iVuucoeG3U1HXjrgzn5FFQ@mail.gmail.com/
Fixes: 79bd37120b1495 ("btrfs: rework chunk allocation to avoid exhaustion of the system chunk array")
CC: stable@vger.kernel.org # 5.14+
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agodma-buf/poll: Get a file reference for outstanding fence callbacks
Michel Dänzer [Fri, 23 Jul 2021 07:58:57 +0000 (09:58 +0200)]
dma-buf/poll: Get a file reference for outstanding fence callbacks

[ Upstream commit ff2d23843f7fb4f13055be5a4a9a20ddd04e6e9c ]

This makes sure we don't hit the

BUG_ON(dmabuf->cb_in.active || dmabuf->cb_out.active);

in dma_buf_release, which could be triggered by user space closing the
dma-buf file description while there are outstanding fence callbacks
from dma_buf_poll.

Cc: stable@vger.kernel.org
Signed-off-by: Michel Dänzer <mdaenzer@redhat.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210723075857.4065-1-michel@daenzer.net
Signed-off-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoInput: goodix - try not to touch the reset-pin on x86/ACPI devices
Hans de Goede [Tue, 7 Dec 2021 07:15:09 +0000 (23:15 -0800)]
Input: goodix - try not to touch the reset-pin on x86/ACPI devices

[ Upstream commit a2fd46cd3dbb83b373ba74f4043f8dae869c65f1 ]

Unless the controller is not responding at boot or after suspend/resume,
the driver never resets the controller on x86/ACPI platforms. The driver
still requesting the reset pin at probe() though in case it needs it.

Until now the driver has always requested the reset pin with GPIOD_IN
as type. The idea being to put the pin in high-impedance mode to save
power until the driver actually wants to issue a reset.

But this means that just requesting the pin can cause issues, since
requesting it in another mode then GPIOD_ASIS may cause the pinctrl
driver to touch the pin settings. We have already had issues before
due to a bug in the pinctrl-cherryview.c driver which has been fixed in
commit 921daeeca91b ("pinctrl: cherryview: Preserve
CHV_PADCTRL1_INVRXTX_TXDATA flag on GPIOs").

And now it turns out that requesting the reset-pin as GPIOD_IN also stops
the touchscreen from working on the GPD P2 max mini-laptop. The behavior
of putting the pin in high-impedance mode relies on there being some
external pull-up to keep it high and there seems to be no pull-up on the
GPD P2 max, causing things to break.

This commit fixes this by requesting the reset pin as is when using
the x86/ACPI code paths to lookup the GPIOs; and by not dropping it
back into input-mode in case the driver does end up issuing a reset
for error-recovery.

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=209061
Fixes: a7d4b171660c ("Input: goodix - add support for getting IRQ + reset GPIOs on Cherry Trail devices")
Cc: stable@vger.kernel.org
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20211206091116.44466-2-hdegoede@redhat.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoInput: goodix - refactor reset handling
Hans de Goede [Tue, 21 Sep 2021 04:08:26 +0000 (21:08 -0700)]
Input: goodix - refactor reset handling

[ Upstream commit 209bda4741f68f102cf2f272227bfc938e387b51 ]

Refactor reset handling a bit, change the main reset handler
into a new goodix_reset_no_int_sync() helper and add a
goodix_reset() wrapper which calls goodix_int_sync()
separately.

Also push the dev_err() call on reset failure into the
goodix_reset_no_int_sync() and goodix_int_sync() functions,
so that we don't need to have separate dev_err() calls in
all their callers.

This is a preparation patch for adding support for controllers
without flash, which need to have their firmware uploaded and
need some other special handling too.

Reviewed-by: Bastien Nocera <hadess@hadess.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20210920150643.155872-4-hdegoede@redhat.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoInput: goodix - add a goodix.h header file
Hans de Goede [Tue, 21 Sep 2021 04:08:18 +0000 (21:08 -0700)]
Input: goodix - add a goodix.h header file

[ Upstream commit a2233cb7b65a017067e2f2703375ecc930a0ab30 ]

Add a goodix.h header file, and move the register definitions,
and struct declarations there and add prototypes for various
helper functions.

This is a preparation patch for adding support for controllers
without flash, which need to have their firmware uploaded and
need some other special handling too.

Since MAINTAINERS needs updating because of this change anyways,
also add myself as co-maintainer.

Reviewed-by: Bastien Nocera <hadess@hadess.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20210920150643.155872-3-hdegoede@redhat.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoInput: goodix - change goodix_i2c_write() len parameter type to int
Hans de Goede [Tue, 21 Sep 2021 04:08:02 +0000 (21:08 -0700)]
Input: goodix - change goodix_i2c_write() len parameter type to int

[ Upstream commit 31ae0102a34ed863c7d32b10e768036324991679 ]

Change the type of the goodix_i2c_write() len parameter to from 'unsigned'
to 'int' to avoid bare use of 'unsigned', changing it to 'int' makes
goodix_i2c_write()' prototype consistent with goodix_i2c_read().

Reviewed-by: Bastien Nocera <hadess@hadess.net>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Link: https://lore.kernel.org/r/20210920150643.155872-2-hdegoede@redhat.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agoInput: cpcap-pwrbutton - handle errors from platform_get_irq()
Tang Bin [Mon, 6 Sep 2021 01:57:32 +0000 (18:57 -0700)]
Input: cpcap-pwrbutton - handle errors from platform_get_irq()

[ Upstream commit 58ae4004b9c4bb040958cf73986b687a5ea4d85d ]

The function cpcap_power_button_probe() does not perform
sufficient error checking after executing platform_get_irq(),
thus fix it.

Signed-off-by: Tang Bin <tangbin@cmss.chinamobile.com>
Link: https://lore.kernel.org/r/20210802121740.8700-1-tangbin@cmss.chinamobile.com
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: fix warning when freeing leaf after subvolume creation failure
Filipe Manana [Mon, 13 Dec 2021 08:45:13 +0000 (08:45 +0000)]
btrfs: fix warning when freeing leaf after subvolume creation failure

[ Upstream commit 212a58fda9b9077e0efc20200a4feb76afacfd95 ]

When creating a subvolume, at ioctl.c:create_subvol(), if we fail to
insert the root item for the new subvolume into the root tree, we can
trigger the following warning:

[78961.741046] WARNING: CPU: 0 PID: 4079814 at fs/btrfs/extent-tree.c:3357 btrfs_free_tree_block+0x2af/0x310 [btrfs]
[78961.743344] Modules linked in:
[78961.749440]  dm_snapshot dm_thin_pool (...)
[78961.773648] CPU: 0 PID: 4079814 Comm: fsstress Not tainted 5.16.0-rc4-btrfs-next-108 #1
[78961.775198] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
[78961.777266] RIP: 0010:btrfs_free_tree_block+0x2af/0x310 [btrfs]
[78961.778398] Code: 17 00 48 85 (...)
[78961.781067] RSP: 0018:ffffaa4001657b28 EFLAGS: 00010202
[78961.781877] RAX: 0000000000000213 RBX: ffff897f8a796910 RCX: 0000000000000000
[78961.782780] RDX: 0000000000000000 RSI: 0000000011004000 RDI: 00000000ffffffff
[78961.783764] RBP: ffff8981f490e800 R08: 0000000000000001 R09: 0000000000000000
[78961.784740] R10: 0000000000000000 R11: 0000000000000001 R12: ffff897fc963fcc8
[78961.785665] R13: 0000000000000001 R14: ffff898063548000 R15: ffff898063548000
[78961.786620] FS:  00007f31283c6b80(0000) GS:ffff8982ace00000(0000) knlGS:0000000000000000
[78961.787717] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[78961.788598] CR2: 00007f31285c3000 CR3: 000000023fcc8003 CR4: 0000000000370ef0
[78961.789568] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[78961.790585] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[78961.791684] Call Trace:
[78961.792082]  <TASK>
[78961.792359]  create_subvol+0x5d1/0x9a0 [btrfs]
[78961.793054]  btrfs_mksubvol+0x447/0x4c0 [btrfs]
[78961.794009]  ? preempt_count_add+0x49/0xa0
[78961.794705]  __btrfs_ioctl_snap_create+0x123/0x190 [btrfs]
[78961.795712]  ? _copy_from_user+0x66/0xa0
[78961.796382]  btrfs_ioctl_snap_create_v2+0xbb/0x140 [btrfs]
[78961.797392]  btrfs_ioctl+0xd1e/0x35c0 [btrfs]
[78961.798172]  ? __slab_free+0x10a/0x360
[78961.798820]  ? rcu_read_lock_sched_held+0x12/0x60
[78961.799664]  ? lock_release+0x223/0x4a0
[78961.800321]  ? lock_acquired+0x19f/0x420
[78961.800992]  ? rcu_read_lock_sched_held+0x12/0x60
[78961.801796]  ? trace_hardirqs_on+0x1b/0xe0
[78961.802495]  ? _raw_spin_unlock_irqrestore+0x3e/0x60
[78961.803358]  ? kmem_cache_free+0x321/0x3c0
[78961.804071]  ? __x64_sys_ioctl+0x83/0xb0
[78961.804711]  __x64_sys_ioctl+0x83/0xb0
[78961.805348]  do_syscall_64+0x3b/0xc0
[78961.805969]  entry_SYSCALL_64_after_hwframe+0x44/0xae
[78961.806830] RIP: 0033:0x7f31284bc957
[78961.807517] Code: 3c 1c 48 f7 d8 (...)

This is because we are calling btrfs_free_tree_block() on an extent
buffer that is dirty. Fix that by cleaning the extent buffer, with
btrfs_clean_tree_block(), before freeing it.

This was triggered by test case generic/475 from fstests.

Fixes: 67addf29004c5b ("btrfs: fix metadata extent leak after failure to create subvolume")
CC: stable@vger.kernel.org # 4.4+
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: fix invalid delayed ref after subvolume creation failure
Filipe Manana [Mon, 13 Dec 2021 08:45:12 +0000 (08:45 +0000)]
btrfs: fix invalid delayed ref after subvolume creation failure

[ Upstream commit 7a1636089acfee7562fe79aff7d1b4c57869896d ]

When creating a subvolume, at ioctl.c:create_subvol(), if we fail to
insert the new root's root item into the root tree, we are freeing the
metadata extent we reserved for the new root to prevent a metadata
extent leak, as we don't abort the transaction at that point (since
there is nothing at that point that is irreversible).

However we allocated the metadata extent for the new root which we are
creating for the new subvolume, so its delayed reference refers to the
ID of this new root. But when we free the metadata extent we pass the
root of the subvolume where the new subvolume is located to
btrfs_free_tree_block() - this is incorrect because this will generate
a delayed reference that refers to the ID of the parent subvolume's root,
and not to ID of the new root.

This results in a failure when running delayed references that leads to
a transaction abort and a trace like the following:

[3868.738042] RIP: 0010:__btrfs_free_extent+0x709/0x950 [btrfs]
[3868.739857] Code: 68 0f 85 e6 fb ff (...)
[3868.742963] RSP: 0018:ffffb0e9045cf910 EFLAGS: 00010246
[3868.743908] RAX: 00000000fffffffe RBX: 00000000fffffffe RCX: 0000000000000002
[3868.745312] RDX: 00000000fffffffe RSI: 0000000000000002 RDI: ffff90b0cd793b88
[3868.746643] RBP: 000000000e5d8000 R08: 0000000000000000 R09: ffff90b0cd793b88
[3868.747979] R10: 0000000000000002 R11: 00014ded97944d68 R12: 0000000000000000
[3868.749373] R13: ffff90b09afe4a28 R14: 0000000000000000 R15: ffff90b0cd793b88
[3868.750725] FS:  00007f281c4a8b80(0000) GS:ffff90b3ada00000(0000) knlGS:0000000000000000
[3868.752275] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[3868.753515] CR2: 00007f281c6a5000 CR3: 0000000108a42006 CR4: 0000000000370ee0
[3868.754869] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[3868.756228] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[3868.757803] Call Trace:
[3868.758281]  <TASK>
[3868.758655]  ? btrfs_merge_delayed_refs+0x178/0x1c0 [btrfs]
[3868.759827]  __btrfs_run_delayed_refs+0x2b1/0x1250 [btrfs]
[3868.761047]  btrfs_run_delayed_refs+0x86/0x210 [btrfs]
[3868.762069]  ? lock_acquired+0x19f/0x420
[3868.762829]  btrfs_commit_transaction+0x69/0xb20 [btrfs]
[3868.763860]  ? _raw_spin_unlock+0x29/0x40
[3868.764614]  ? btrfs_block_rsv_release+0x1c2/0x1e0 [btrfs]
[3868.765870]  create_subvol+0x1d8/0x9a0 [btrfs]
[3868.766766]  btrfs_mksubvol+0x447/0x4c0 [btrfs]
[3868.767669]  ? preempt_count_add+0x49/0xa0
[3868.768444]  __btrfs_ioctl_snap_create+0x123/0x190 [btrfs]
[3868.769639]  ? _copy_from_user+0x66/0xa0
[3868.770391]  btrfs_ioctl_snap_create_v2+0xbb/0x140 [btrfs]
[3868.771495]  btrfs_ioctl+0xd1e/0x35c0 [btrfs]
[3868.772364]  ? __slab_free+0x10a/0x360
[3868.773198]  ? rcu_read_lock_sched_held+0x12/0x60
[3868.774121]  ? lock_release+0x223/0x4a0
[3868.774863]  ? lock_acquired+0x19f/0x420
[3868.775634]  ? rcu_read_lock_sched_held+0x12/0x60
[3868.776530]  ? trace_hardirqs_on+0x1b/0xe0
[3868.777373]  ? _raw_spin_unlock_irqrestore+0x3e/0x60
[3868.778280]  ? kmem_cache_free+0x321/0x3c0
[3868.779011]  ? __x64_sys_ioctl+0x83/0xb0
[3868.779718]  __x64_sys_ioctl+0x83/0xb0
[3868.780387]  do_syscall_64+0x3b/0xc0
[3868.781059]  entry_SYSCALL_64_after_hwframe+0x44/0xae
[3868.781953] RIP: 0033:0x7f281c59e957
[3868.782585] Code: 3c 1c 48 f7 d8 4c (...)
[3868.785867] RSP: 002b:00007ffe1f83e2b8 EFLAGS: 00000202 ORIG_RAX: 0000000000000010
[3868.787198] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f281c59e957
[3868.788450] RDX: 00007ffe1f83e2c0 RSI: 0000000050009418 RDI: 0000000000000003
[3868.789748] RBP: 00007ffe1f83f300 R08: 0000000000000000 R09: 00007ffe1f83fe36
[3868.791214] R10: 0000000000000000 R11: 0000000000000202 R12: 0000000000000003
[3868.792468] R13: 0000000000000003 R14: 00007ffe1f83e2c0 R15: 00000000000003cc
[3868.793765]  </TASK>
[3868.794037] irq event stamp: 0
[3868.794548] hardirqs last  enabled at (0): [<0000000000000000>] 0x0
[3868.795670] hardirqs last disabled at (0): [<ffffffff98294214>] copy_process+0x934/0x2040
[3868.797086] softirqs last  enabled at (0): [<ffffffff98294214>] copy_process+0x934/0x2040
[3868.798309] softirqs last disabled at (0): [<0000000000000000>] 0x0
[3868.799284] ---[ end trace be24c7002fe27747 ]---
[3868.799928] BTRFS info (device dm-0): leaf 241188864 gen 1268 total ptrs 214 free space 469 owner 2
[3868.801133] BTRFS info (device dm-0): refs 2 lock_owner 225627 current 225627
[3868.802056]  item 0 key (237436928 169 0) itemoff 16250 itemsize 33
[3868.802863]          extent refs 1 gen 1265 flags 2
[3868.803447]          ref#0: tree block backref root 1610
(...)
[3869.064354]  item 114 key (241008640 169 0) itemoff 12488 itemsize 33
[3869.065421]          extent refs 1 gen 1268 flags 2
[3869.066115]          ref#0: tree block backref root 1689
(...)
[3869.403834] BTRFS error (device dm-0): unable to find ref byte nr 241008640 parent 0 root 1622  owner 0 offset 0
[3869.405641] BTRFS: error (device dm-0) in __btrfs_free_extent:3076: errno=-2 No such entry
[3869.407138] BTRFS: error (device dm-0) in btrfs_run_delayed_refs:2159: errno=-2 No such entry

Fix this by passing the new subvolume's root ID to btrfs_free_tree_block().
This requires changing the root argument of btrfs_free_tree_block() from
struct btrfs_root * to a u64, since at this point during the subvolume
creation we have not yet created the struct btrfs_root for the new
subvolume, and btrfs_free_tree_block() only needs a root ID and nothing
else from a struct btrfs_root.

This was triggered by test case generic/475 from fstests.

Fixes: 67addf29004c5b ("btrfs: fix metadata extent leak after failure to create subvolume")
CC: stable@vger.kernel.org # 4.4+
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: add additional parameters to btrfs_init_tree_ref/btrfs_init_data_ref
Nikolay Borisov [Tue, 12 Oct 2021 08:21:35 +0000 (11:21 +0300)]
btrfs: add additional parameters to btrfs_init_tree_ref/btrfs_init_data_ref

[ Upstream commit f42c5da6c12e990d8ec415199600b4d593c63bf5 ]

In order to make 'real_root' used only in ref-verify it's required to
have the necessary context to perform the same checks that this member
is used for. So add 'mod_root' which will contain the root on behalf of
which a delayed ref was created and a 'skip_group' parameter which
will contain callsite-specific override of skip_qgroup.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agobtrfs: rename btrfs_alloc_chunk to btrfs_create_chunk
Nikolay Borisov [Wed, 18 Aug 2021 10:41:19 +0000 (13:41 +0300)]
btrfs: rename btrfs_alloc_chunk to btrfs_create_chunk

[ Upstream commit f6f39f7a0add4e7fd120a709545b57586a1d0393 ]

The user facing function used to allocate new chunks is
btrfs_chunk_alloc, unfortunately there is yet another similar sounding
function - btrfs_alloc_chunk. This creates confusion, especially since
the latter function can be considered "private" in the sense that it
implements the first stage of chunk creation and as such is called by
btrfs_chunk_alloc.

To avoid the awkwardness that comes with having similarly named but
distinctly different in their purpose function rename btrfs_alloc_chunk
to btrfs_create_chunk, given that the main purpose of this function is
to orchestrate the whole process of allocating a chunk - reserving space
into devices, deciding on characteristics of the stripe size and
creating the in-memory structures.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
2 years agonetfilter: nf_tables: stricter validation of element data
Pablo Neira Ayuso [Sat, 2 Jul 2022 02:16:30 +0000 (04:16 +0200)]
netfilter: nf_tables: stricter validation of element data

commit 7e6bc1f6cabcd30aba0b11219d8e01b952eacbb6 upstream.

Make sure element data type and length do not mismatch the one specified
by the set declaration.

Fixes: 7d7402642eaf ("netfilter: nf_tables: variable sized set element keys / data")
Reported-by: Hugues ANGUELKOV <hanguelkov@randorisec.fr>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agonetfilter: nft_set_pipapo: release elements in clone from abort path
Pablo Neira Ayuso [Sat, 2 Jul 2022 02:16:31 +0000 (04:16 +0200)]
netfilter: nft_set_pipapo: release elements in clone from abort path

commit 9827a0e6e23bf43003cd3d5b7fb11baf59a35e1e upstream.

New elements that reside in the clone are not released in case that the
transaction is aborted.

[16302.231754] ------------[ cut here ]------------
[16302.231756] WARNING: CPU: 0 PID: 100509 at net/netfilter/nf_tables_api.c:1864 nf_tables_chain_destroy+0x26/0x127 [nf_tables]
[...]
[16302.231882] CPU: 0 PID: 100509 Comm: nft Tainted: G        W         5.19.0-rc3+ #155
[...]
[16302.231887] RIP: 0010:nf_tables_chain_destroy+0x26/0x127 [nf_tables]
[16302.231899] Code: f3 fe ff ff 41 55 41 54 55 53 48 8b 6f 10 48 89 fb 48 c7 c7 82 96 d9 a0 8b 55 50 48 8b 75 58 e8 de f5 92 e0 83 7d 50 00 74 09 <0f> 0b 5b 5d 41 5c 41 5d c3 4c 8b 65 00 48 8b 7d 08 49 39 fc 74 05
[...]
[16302.231917] Call Trace:
[16302.231919]  <TASK>
[16302.231921]  __nf_tables_abort.cold+0x23/0x28 [nf_tables]
[16302.231934]  nf_tables_abort+0x30/0x50 [nf_tables]
[16302.231946]  nfnetlink_rcv_batch+0x41a/0x840 [nfnetlink]
[16302.231952]  ? __nla_validate_parse+0x48/0x190
[16302.231959]  nfnetlink_rcv+0x110/0x129 [nfnetlink]
[16302.231963]  netlink_unicast+0x211/0x340
[16302.231969]  netlink_sendmsg+0x21e/0x460

Add nft_set_pipapo_match_destroy() helper function to release the
elements in the lookup tables.

Stefano Brivio says: "We additionally look for elements pointers in the
cloned matching data if priv->dirty is set, because that means that
cloned data might point to additional elements we did not commit to the
working copy yet (such as the abort path case, but perhaps not limited
to it)."

Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges")
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agonet: rose: fix UAF bug caused by rose_t0timer_expiry
Duoming Zhou [Tue, 5 Jul 2022 12:56:10 +0000 (20:56 +0800)]
net: rose: fix UAF bug caused by rose_t0timer_expiry

commit 148ca04518070910739dfc4eeda765057856403d upstream.

There are UAF bugs caused by rose_t0timer_expiry(). The
root cause is that del_timer() could not stop the timer
handler that is running and there is no synchronization.
One of the race conditions is shown below:

    (thread 1)             |        (thread 2)
                           | rose_device_event
                           |   rose_rt_device_down
                           |     rose_remove_neigh
rose_t0timer_expiry        |       rose_stop_t0timer(rose_neigh)
  ...                      |         del_timer(&neigh->t0timer)
                           |         kfree(rose_neigh) //[1]FREE
  neigh->dce_mode //[2]USE |

The rose_neigh is deallocated in position [1] and use in
position [2].

The crash trace triggered by POC is like below:

BUG: KASAN: use-after-free in expire_timers+0x144/0x320
Write of size 8 at addr ffff888009b19658 by task swapper/0/0
...
Call Trace:
 <IRQ>
 dump_stack_lvl+0xbf/0xee
 print_address_description+0x7b/0x440
 print_report+0x101/0x230
 ? expire_timers+0x144/0x320
 kasan_report+0xed/0x120
 ? expire_timers+0x144/0x320
 expire_timers+0x144/0x320
 __run_timers+0x3ff/0x4d0
 run_timer_softirq+0x41/0x80
 __do_softirq+0x233/0x544
 ...

This patch changes rose_stop_ftimer() and rose_stop_t0timer()
in rose_remove_neigh() to del_timer_sync() in order that the
timer handler could be finished before the resources such as
rose_neigh and so on are deallocated. As a result, the UAF
bugs could be mitigated.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Link: https://lore.kernel.org/r/20220705125610.77971-1-duoming@zju.edu.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agousbnet: fix memory leak in error case
Oliver Neukum [Tue, 5 Jul 2022 12:53:51 +0000 (14:53 +0200)]
usbnet: fix memory leak in error case

commit b55a21b764c1e182014630fa5486d717484ac58f upstream.

usbnet_write_cmd_async() mixed up which buffers
need to be freed in which error case.

v2: add Fixes tag
v3: fix uninitialized buf pointer

Fixes: 877bd862f32b8 ("usbnet: introduce usbnet 3 command helpers")
Signed-off-by: Oliver Neukum <oneukum@suse.com>
Link: https://lore.kernel.org/r/20220705125351.17309-1-oneukum@suse.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
2 years agobpf: Fix insufficient bounds propagation from adjust_scalar_min_max_vals
Daniel Borkmann [Fri, 1 Jul 2022 12:47:25 +0000 (14:47 +0200)]
bpf: Fix insufficient bounds propagation from adjust_scalar_min_max_vals

commit 3844d153a41adea718202c10ae91dc96b37453b5 upstream.

Kuee reported a corner case where the tnum becomes constant after the call
to __reg_bound_offset(), but the register's bounds are not, that is, its
min bounds are still not equal to the register's max bounds.

This in turn allows to leak pointers through turning a pointer register as
is into an unknown scalar via adjust_ptr_min_max_vals().

Before:

  func#0 @0
  0: R1=ctx(off=0,imm=0,umax=0,var_off=(0x0; 0x0)) R10=fp(off=0,imm=0,umax=0,var_off=(0x0; 0x0))
  0: (b7) r0 = 1                        ; R0_w=scalar(imm=1,umin=1,umax=1,var_off=(0x1; 0x0))
  1: (b7) r3 = 0                        ; R3_w=scalar(imm=0,umax=0,var_off=(0x0; 0x0))
  2: (87) r3 = -r3                      ; R3_w=scalar()
  3: (87) r3 = -r3                      ; R3_w=scalar()
  4: (47) r3 |= 32767                   ; R3_w=scalar(smin=-9223372036854743041,umin=32767,var_off=(0x7fff; 0xffffffffffff8000),s32_min=-2147450881)
  5: (75) if r3 s>= 0x0 goto pc+1       ; R3_w=scalar(umin=9223372036854808575,var_off=(0x8000000000007fff; 0x7fffffffffff8000),s32_min=-2147450881,u32_min=32767)
  6: (95) exit

  from 5 to 7: R0=scalar(imm=1,umin=1,umax=1,var_off=(0x1; 0x0)) R1=ctx(off=0,imm=0,umax=0,var_off=(0x0; 0x0)) R3=scalar(umin=32767,umax=9223372036854775807,var_off=(0x7fff; 0x7fffffffffff8000),s32_min=-2147450881) R10=fp(off=0,imm=0,umax=0,var_off=(0x0; 0x0))
  7: (d5) if r3 s<= 0x8000 goto pc+1    ; R3=scalar(umin=32769,umax=9223372036854775807,var_off=(0x7fff; 0x7fffffffffff8000),s32_min=-2147450881,u32_min=32767)
  8: (95) exit

  from 7 to 9: R0=scalar(imm=1,umin=1,umax=1,var_off=(0x1; 0x0)) R1=ctx(off=0,imm=0,umax=0,var_off=(0x0; 0x0)) R3=scalar(umin=32767,umax=32768,var_off=(0x7fff; 0x8000)) R10=fp(off=0,imm=0,umax=0,var_off=(0x0; 0x0))
  9: (07) r3 += -32767                  ; R3_w=scalar(imm=0,umax=1,var_off=(0x0; 0x0))  <--- [*]
  10: (95) exit

What can be seen here is that R3=scalar(umin=32767,umax=32768,var_off=(0x7fff;
0x8000)) after the operation R3 += -32767 results in a 'malformed' constant, that
is, R3_w=scalar(imm=0,umax=1,var_off=(0x0; 0x0)). Intersecting with var_off has
not been done at that point via __update_reg_bounds(), which would have improved
the umax to be equal to umin.

Refactor the tnum <> min/max bounds information flow into a reg_bounds_sync()
helper and use it consistently everywhere. After the fix, bounds have been
corrected to R3_w=scalar(imm=0,umax=0,var_off=(0x0; 0x0)) and thus the register
is regarded as a 'proper' constant scalar of 0.

After:

  func#0 @0
  0: R1=ctx(off=0,imm=0,umax=0,var_off=(0x0; 0x0)) R10=fp(off=0,imm=0,umax=0,var_off=(0x0; 0x0))
  0: (b7) r0 = 1                        ; R0_w=scalar(imm=1,umin=1,umax=1,var_off=(0x1; 0x0))
  1: (b7) r3 = 0                        ; R3_w=scalar(imm=0,umax=0,var_off=(0x0; 0x0))
  2: (87) r3 = -r3                      ; R3_w=scalar()
  3: (87) r3 = -r3                      ; R3_w=scalar()
  4: (47) r3 |= 32767                   ; R3_w=scalar(smin=-9223372036854743041,umin=32767,var_off=(0x7fff; 0xffffffffffff8000),s32_min=-2147450881)
  5: (75) if r3 s>= 0x0 goto pc+1       ; R3_w=scalar(umin=9223372036854808575,var_off=(0x8000000000007fff; 0x7fffffffffff8000),s32_min=-2147450881,u32_min=32767)
  6: (95) exit

  from 5 to 7: R0=scalar(imm=1,umin=1,umax=1,var_off=(0x1; 0x0)) R1=ctx(off=0,imm=0,umax=0,var_off=(0x0; 0x0)) R3=scalar(umin=32767,umax=9223372036854775807,var_off=(0x7fff; 0x7fffffffffff8000),s32_min=-2147450881) R10=fp(off=0,imm=0,umax=0,var_off=(0x0; 0x0))
  7: (d5) if r3 s<= 0x8000 goto pc+1    ; R3=scalar(umin=32769,umax=9223372036854775807,var_off=(0x7fff; 0x7fffffffffff8000),s32_min=-2147450881,u32_min=32767)
  8: (95) exit

  from 7 to 9: R0=scalar(imm=1,umin=1,umax=1,var_off=(0x1; 0x0)) R1=ctx(off=0,imm=0,umax=0,var_off=(0x0; 0x0)) R3=scalar(umin=32767,umax=32768,var_off=(0x7fff; 0x8000)) R10=fp(off=0,imm=0,umax=0,var_off=(0x0; 0x0))
  9: (07) r3 += -32767                  ; R3_w=scalar(imm=0,umax=0,var_off=(0x0; 0x0))  <--- [*]
  10: (95) exit

Fixes: b03c9f9fdc37 ("bpf/verifier: track signed and unsigned min/max values")
Reported-by: Kuee K1r0a <liulin063@gmail.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Andrii Nakryiko <andrii@kernel.org>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Link: https://lore.kernel.org/bpf/20220701124727.11153-2-daniel@iogearbox.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>