platform/kernel/linux-rpi.git
4 years agoKVM: arm64: Stop writing aarch32's CSSELR into ACTLR
James Morse [Fri, 29 May 2020 15:06:54 +0000 (15:06 +0000)]
KVM: arm64: Stop writing aarch32's CSSELR into ACTLR

commit 7c582bf4ed84f3eb58bdd1f63024a14c17551e7d upstream.

aarch32 has pairs of registers to access the high and low parts of 64bit
registers. KVM has a union of 64bit sys_regs[] and 32bit copro[]. The
32bit accessors read the high or low part of the 64bit sys_reg[] value
through the union.

Both sys_reg_descs[] and cp15_regs[] list access_csselr() as the accessor
for CSSELR{,_EL1}. access_csselr() is only aware of the 64bit sys_regs[],
and expects r->reg to be 'CSSELR_EL1' in the enum, index 2 of the 64bit
array.

cp15_regs[] uses the 32bit copro[] alias of sys_regs[]. Here CSSELR is
c0_CSSELR which is the same location in sys_reg[]. r->reg is 'c0_CSSELR',
index 4 in the 32bit array.

access_csselr() uses the 32bit r->reg value to access the 64bit array,
so reads and write the wrong value. sys_regs[4], is ACTLR_EL1, which
is subsequently save/restored when we enter the guest.

ACTLR_EL1 is supposed to be read-only for the guest. This register
only affects execution at EL1, and the host's value is restored before
we return to host EL1.

Convert the 32bit register index back to the 64bit version.

Suggested-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: James Morse <james.morse@arm.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200529150656.7339-2-james.morse@arm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: MIPS: Fix VPN2_MASK definition for variable cpu_vmbits
Xing Li [Sat, 23 May 2020 07:56:29 +0000 (15:56 +0800)]
KVM: MIPS: Fix VPN2_MASK definition for variable cpu_vmbits

commit 5816c76dea116a458f1932eefe064e35403248eb upstream.

If a CPU support more than 32bit vmbits (which is true for 64bit CPUs),
VPN2_MASK set to fixed 0xffffe000 will lead to a wrong EntryHi in some
functions such as _kvm_mips_host_tlb_inv().

The cpu_vmbits definition of 32bit CPU in cpu-features.h is 31, so we
still use the old definition.

Cc: Stable <stable@vger.kernel.org>
Reviewed-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Signed-off-by: Xing Li <lixing@loongson.cn>
[Huacai: Improve commit messages]
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Message-Id: <1590220602-3547-3-git-send-email-chenhc@lemote.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: MIPS: Define KVM_ENTRYHI_ASID to cpu_asid_mask(&boot_cpu_data)
Xing Li [Sat, 23 May 2020 07:56:28 +0000 (15:56 +0800)]
KVM: MIPS: Define KVM_ENTRYHI_ASID to cpu_asid_mask(&boot_cpu_data)

commit fe2b73dba47fb6d6922df1ad44e83b1754d5ed4d upstream.

The code in decode_config4() of arch/mips/kernel/cpu-probe.c

        asid_mask = MIPS_ENTRYHI_ASID;
        if (config4 & MIPS_CONF4_AE)
                asid_mask |= MIPS_ENTRYHI_ASIDX;
        set_cpu_asid_mask(c, asid_mask);

set asid_mask to cpuinfo->asid_mask.

So in order to support variable ASID_MASK, KVM_ENTRYHI_ASID should also
be changed to cpu_asid_mask(&boot_cpu_data).

Cc: Stable <stable@vger.kernel.org> #4.9+
Reviewed-by: Aleksandar Markovic <aleksandar.qemu.devel@gmail.com>
Signed-off-by: Xing Li <lixing@loongson.cn>
[Huacai: Change current_cpu_data to boot_cpu_data for optimization]
Signed-off-by: Huacai Chen <chenhc@lemote.com>
Message-Id: <1590220602-3547-2-git-send-email-chenhc@lemote.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: nVMX: Consult only the "basic" exit reason when routing nested exit
Sean Christopherson [Thu, 27 Feb 2020 17:44:30 +0000 (09:44 -0800)]
KVM: nVMX: Consult only the "basic" exit reason when routing nested exit

commit 2ebac8bb3c2d35f5135466490fc8eeaf3f3e2d37 upstream.

Consult only the basic exit reason, i.e. bits 15:0 of vmcs.EXIT_REASON,
when determining whether a nested VM-Exit should be reflected into L1 or
handled by KVM in L0.

For better or worse, the switch statement in nested_vmx_exit_reflected()
currently defaults to "true", i.e. reflects any nested VM-Exit without
dedicated logic.  Because the case statements only contain the basic
exit reason, any VM-Exit with modifier bits set will be reflected to L1,
even if KVM intended to handle it in L0.

Practically speaking, this only affects EXIT_REASON_MCE_DURING_VMENTRY,
i.e. a #MC that occurs on nested VM-Enter would be incorrectly routed to
L1, as "failed VM-Entry" is the only modifier that KVM can currently
encounter.  The SMM modifiers will never be generated as KVM doesn't
support/employ a SMI Transfer Monitor.  Ditto for "exit from enclave",
as KVM doesn't yet support virtualizing SGX, i.e. it's impossible to
enter an enclave in a KVM guest (L1 or L2).

Fixes: 644d711aa0e1 ("KVM: nVMX: Deciding if L0 or L1 should handle an L2 exit")
Cc: Jim Mattson <jmattson@google.com>
Cc: Xiaoyao Li <xiaoyao.li@intel.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20200227174430.26371-1-sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: nSVM: leave ASID aside in copy_vmcb_control_area
Paolo Bonzini [Wed, 20 May 2020 12:02:17 +0000 (08:02 -0400)]
KVM: nSVM: leave ASID aside in copy_vmcb_control_area

commit 6c0238c4a62b3a0b1201aeb7e33a4636d552a436 upstream.

Restoring the ASID from the hsave area on VMEXIT is wrong, because its
value depends on the handling of TLB flushes.  Just skipping the field in
copy_vmcb_control_area will do.

Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: nSVM: fix condition for filtering async PF
Paolo Bonzini [Sat, 16 May 2020 13:19:06 +0000 (09:19 -0400)]
KVM: nSVM: fix condition for filtering async PF

commit a3535be731c2a343912578465021f50937f7b099 upstream.

Async page faults have to be trapped in the host (L1 in this case),
since the APF reason was passed from L0 to L1 and stored in the L1 APF
data page.  This was completely reversed: the page faults were passed
to the guest, a L2 hypervisor.

Cc: stable@vger.kernel.org
Reviewed-by: Sean Christopherson <sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: nVMX: Skip IBPB when switching between vmcs01 and vmcs02
Sean Christopherson [Fri, 1 May 2020 16:31:17 +0000 (09:31 -0700)]
KVM: nVMX: Skip IBPB when switching between vmcs01 and vmcs02

commit 5c911beff20aa8639e7a1f28988736c13e03ed54 upstream.

Skip the Indirect Branch Prediction Barrier that is triggered on a VMCS
switch when running with spectre_v2_user=on/auto if the switch is
between two VMCSes in the same guest, i.e. between vmcs01 and vmcs02.
The IBPB is intended to prevent one guest from attacking another, which
is unnecessary in the nested case as it's the same guest from KVM's
perspective.

This all but eliminates the overhead observed for nested VMX transitions
when running with CONFIG_RETPOLINE=y and spectre_v2_user=on/auto, which
can be significant, e.g. roughly 3x on current systems.

Reported-by: Alexander Graf <graf@amazon.com>
Cc: KarimAllah Raslan <karahmed@amazon.de>
Cc: stable@vger.kernel.org
Fixes: 15d45071523d ("KVM/x86: Add IBPB support")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20200501163117.4655-1-sean.j.christopherson@intel.com>
[Invert direction of bool argument. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agovideo: fbdev: w100fb: Fix a potential double free.
Christophe JAILLET [Wed, 6 May 2020 18:19:02 +0000 (20:19 +0200)]
video: fbdev: w100fb: Fix a potential double free.

commit 18722d48a6bb9c2e8d046214c0a5fd19d0a7c9f6 upstream.

Some memory is vmalloc'ed in the 'w100fb_save_vidmem' function and freed in
the 'w100fb_restore_vidmem' function. (these functions are called
respectively from the 'suspend' and the 'resume' functions)

However, it is also freed in the 'remove' function.

In order to avoid a potential double free, set the corresponding pointer
to NULL once freed in the 'w100fb_restore_vidmem' function.

Fixes: aac51f09d96a ("[PATCH] w100fb: Rewrite for platform independence")
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Antonino Daplas <adaplas@pol.net>
Cc: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Cc: <stable@vger.kernel.org> # v2.6.14+
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20200506181902.193290-1-christophe.jaillet@wanadoo.fr
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agovideo: vt8500lcdfb: fix fallthrough warning
Sam Ravnborg [Sun, 12 Apr 2020 20:21:43 +0000 (22:21 +0200)]
video: vt8500lcdfb: fix fallthrough warning

commit 1c49f35e9e9156273124a0cfd38b57f7a7d4828f upstream.

Fix following warning:
vt8500lcdfb.c: In function 'vt8500lcd_blank':
vt8500lcdfb.c:229:6: warning: this statement may fall through [-Wimplicit-fallthrough=]
      if (info->fix.visual == FB_VISUAL_PSEUDOCOLOR ||
         ^
vt8500lcdfb.c:233:2: note: here
     case FB_BLANK_UNBLANK:
     ^~~~

Adding a simple "fallthrough;" fixed the warning.
The fix was build tested.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
Reported-by: kbuild test robot <lkp@intel.com>
Fixes: e41f1a989408 ("fbdev: Implement simple blanking in pseudocolor modes for vt8500lcdfb")
Cc: Alexey Charkov <alchark@gmail.com>
Cc: Paul Mundt <lethal@linux-sh.org>
Cc: <stable@vger.kernel.org> # v2.6.38+
Signed-off-by: Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200412202143.GA26948@ravnborg.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoEDAC/skx: Use the mcmtr register to retrieve close_pg/bank_xor_enable
Qiuxu Zhuo [Fri, 15 May 2020 12:34:06 +0000 (20:34 +0800)]
EDAC/skx: Use the mcmtr register to retrieve close_pg/bank_xor_enable

commit 1032095053b34d474aa20f2625d97dd306e0991b upstream.

The skx_edac driver wrongly uses the mtr register to retrieve two fields
close_pg and bank_xor_enable. Fix it by using the correct mcmtr register
to get the two fields.

Cc: <stable@vger.kernel.org>
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Reported-by: Matthew Riley <mattdr@google.com>
Acked-by: Aristeu Rozanski <aris@redhat.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Link: https://lore.kernel.org/r/20200515210146.1337-1-tony.luck@intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agocpufreq: Fix up cpufreq_boost_set_sw()
Rafael J. Wysocki [Mon, 18 May 2020 10:49:45 +0000 (12:49 +0200)]
cpufreq: Fix up cpufreq_boost_set_sw()

commit 552abb884e97d26589964e5a8c7e736f852f95f0 upstream.

After commit 18c49926c4bf ("cpufreq: Add QoS requests for userspace
constraints") the return value of freq_qos_update_request(), that can
be 1, passed by cpufreq_boost_set_sw() to its caller sometimes
confuses the latter, which only expects to see 0 or negative error
codes, so notice that cpufreq_boost_set_sw() can return an error code
(which should not be -EINVAL for that matter) as soon as the first
policy without a frequency table is found (because either all policies
have a frequency table or none of them have it) and rework it to meet
its caller's expectations.

Fixes: 18c49926c4bf ("cpufreq: Add QoS requests for userspace constraints")
Reported-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Reported-by: Xiongfeng Wang <wangxiongfeng2@huawei.com>
Acked-by: Viresh Kumar <viresh.kumar@linaro.org>
Cc: 5.3+ <stable@vger.kernel.org> # 5.3+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoremoteproc: Fix and restore the parenting hierarchy for vdev
Suman Anna [Mon, 20 Apr 2020 16:06:00 +0000 (11:06 -0500)]
remoteproc: Fix and restore the parenting hierarchy for vdev

commit c774ad010873bb89dcc0cdcb1e96aef6664d8caf upstream.

The commit 086d08725d34 ("remoteproc: create vdev subdevice with specific
dma memory pool") has introduced a new vdev subdevice for each vdev
declared in the firmware resource table and made it as the parent for the
created virtio rpmsg devices instead of the previous remoteproc device.
This changed the overall parenting hierarchy for the rpmsg devices, which
were children of virtio devices, and does not allow the corresponding
rpmsg drivers to retrieve the parent rproc device through the
rproc_get_by_child() API.

Fix this by restoring the remoteproc device as the parent. The new vdev
subdevice can continue to inherit the DMA attributes from the remoteproc's
parent device (actual platform device).

Cc: stable@vger.kernel.org
Fixes: 086d08725d34 ("remoteproc: create vdev subdevice with specific dma memory pool")
Signed-off-by: Suman Anna <s-anna@ti.com>
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
Link: https://lore.kernel.org/r/20200420160600.10467-3-s-anna@ti.com
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoremoteproc: Fall back to using parent memory pool if no dedicated available
Tero Kristo [Mon, 20 Apr 2020 16:05:59 +0000 (11:05 -0500)]
remoteproc: Fall back to using parent memory pool if no dedicated available

commit db9178a4f8c4e523f824892cb8bab00961b07385 upstream.

In some cases, like with OMAP remoteproc, we are not creating dedicated
memory pool for the virtio device. Instead, we use the same memory pool
for all shared memories. The current virtio memory pool handling forces
a split between these two, as a separate device is created for it,
causing memory to be allocated from bad location if the dedicated pool
is not available. Fix this by falling back to using the parent device
memory pool if dedicated is not available.

Cc: stable@vger.kernel.org
Reviewed-by: Mathieu Poirier <mathieu.poirier@linaro.org>
Acked-by: Arnaud Pouliquen <arnaud.pouliquen@st.com>
Fixes: 086d08725d34 ("remoteproc: create vdev subdevice with specific dma memory pool")
Signed-off-by: Tero Kristo <t-kristo@ti.com>
Signed-off-by: Suman Anna <s-anna@ti.com>
Link: https://lore.kernel.org/r/20200420160600.10467-2-s-anna@ti.com
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoproc: Use new_inode not new_inode_pseudo
Eric W. Biederman [Fri, 12 Jun 2020 14:42:03 +0000 (09:42 -0500)]
proc: Use new_inode not new_inode_pseudo

commit ef1548adada51a2f32ed7faef50aa465e1b4c5da upstream.

Recently syzbot reported that unmounting proc when there is an ongoing
inotify watch on the root directory of proc could result in a use
after free when the watch is removed after the unmount of proc
when the watcher exits.

Commit 69879c01a0c3 ("proc: Remove the now unnecessary internal mount
of proc") made it easier to unmount proc and allowed syzbot to see the
problem, but looking at the code it has been around for a long time.

Looking at the code the fsnotify watch should have been removed by
fsnotify_sb_delete in generic_shutdown_super.  Unfortunately the inode
was allocated with new_inode_pseudo instead of new_inode so the inode
was not on the sb->s_inodes list.  Which prevented
fsnotify_unmount_inodes from finding the inode and removing the watch
as well as made it so the "VFS: Busy inodes after unmount" warning
could not find the inodes to warn about them.

Make all of the inodes in proc visible to generic_shutdown_super,
and fsnotify_sb_delete by using new_inode instead of new_inode_pseudo.
The only functional difference is that new_inode places the inodes
on the sb->s_inodes list.

I wrote a small test program and I can verify that without changes it
can trigger this issue, and by replacing new_inode_pseudo with
new_inode the issues goes away.

Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/000000000000d788c905a7dfa3f4@google.com
Reported-by: syzbot+7d2debdcdb3cb93c1e5e@syzkaller.appspotmail.com
Fixes: 0097875bd415 ("proc: Implement /proc/thread-self to point at the directory of the current thread")
Fixes: 021ada7dff22 ("procfs: switch /proc/self away from proc_dir_entry")
Fixes: 51f0885e5415 ("vfs,proc: guarantee unique inodes in /proc")
Signed-off-by: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoovl: initialize error in ovl_copy_xattr
Yuxuan Shui [Wed, 27 May 2020 03:08:02 +0000 (04:08 +0100)]
ovl: initialize error in ovl_copy_xattr

commit 520da69d265a91c6536c63851cbb8a53946974f0 upstream.

In ovl_copy_xattr, if all the xattrs to be copied are overlayfs private
xattrs, the copy loop will terminate without assigning anything to the
error variable, thus returning an uninitialized value.

If ovl_copy_xattr is called from ovl_clear_empty, this uninitialized error
value is put into a pointer by ERR_PTR(), causing potential invalid memory
accesses down the line.

This commit initialize error with 0. This is the correct value because when
there's no xattr to copy, because all xattrs are private, ovl_copy_xattr
should succeed.

This bug is discovered with the help of INIT_STACK_ALL and clang.

Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>
Link: https://bugs.chromium.org/p/chromium/issues/detail?id=1050405
Fixes: 0956254a2d5b ("ovl: don't copy up opaqueness")
Cc: stable@vger.kernel.org # v4.8
Signed-off-by: Alexander Potapenko <glider@google.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonet/mlx5e: Fix repeated XSK usage on one channel
Maxim Mikityanskiy [Mon, 1 Jun 2020 13:03:44 +0000 (16:03 +0300)]
net/mlx5e: Fix repeated XSK usage on one channel

[ Upstream commit 36d45fb9d2fdf348d778bfe73f0427db1c6f9bc7 ]

After an XSK is closed, the relevant structures in the channel are not
zeroed. If an XSK is opened the second time on the same channel without
recreating channels, the stray values in the structures will lead to
incorrect operation of queues, which causes CQE errors, and the new
socket doesn't work at all.

This patch fixes the issue by explicitly zeroing XSK-related structs in
the channel on XSK close. Note that those structs are zeroed on channel
creation, and usually a configuration change (XDP program is set)
happens on XSK open, which leads to recreating channels, so typical XSK
usecases don't suffer from this issue. However, if XSKs are opened and
closed on the same channel without removing the XDP program, this bug
reproduces.

Fixes: db05815b36cb ("net/mlx5e: Add XSK zero-copy support")
Signed-off-by: Maxim Mikityanskiy <maximmi@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonet/mlx5: Fix fatal error handling during device load
Shay Drory [Thu, 7 May 2020 06:32:53 +0000 (09:32 +0300)]
net/mlx5: Fix fatal error handling during device load

[ Upstream commit b6e0b6bebe0732d5cac51f0791f269d2413b8980 ]

Currently, in case of fatal error during mlx5_load_one(), we cannot
enter error state until mlx5_load_one() is finished, what can take
several minutes until commands will get timeouts, because these commands
can't be processed due to the fatal error.
Fix it by setting dev->state as MLX5_DEVICE_STATE_INTERNAL_ERROR before
requesting the lock.

Fixes: c1d4d2e92ad6 ("net/mlx5: Avoid calling sleeping function by the health poll thread")
Signed-off-by: Shay Drory <shayd@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonet/mlx5: drain health workqueue in case of driver load error
Shay Drory [Wed, 6 May 2020 12:59:48 +0000 (15:59 +0300)]
net/mlx5: drain health workqueue in case of driver load error

[ Upstream commit 42ea9f1b5c625fad225d4ac96a7e757dd4199d9c ]

In case there is a work in the health WQ when we teardown the driver,
in driver load error flow, the health work will try to read dev->iseg,
which was already unmap in mlx5_pci_close().
Fix it by draining the health workqueue first thing in mlx5_pci_close().

Trace of the error:
BUG: unable to handle page fault for address: ffffb5b141c18014
PF: supervisor read access in kernel mode
PF: error_code(0x0000) - not-present page
PGD 1fe95d067 P4D 1fe95d067 PUD 1fe95e067 PMD 1b7823067 PTE 0
Oops: 0000 [#1] SMP PTI
CPU: 3 PID: 6755 Comm: kworker/u128:2 Not tainted 5.2.0-net-next-mlx5-hv_stats-over-last-worked-hyperv #1
Hardware name: Microsoft Corporation Virtual Machine/Virtual Machine, BIOS 090006  04/28/2016
Workqueue: mlx5_healtha050:00:02.0 mlx5_fw_fatal_reporter_err_work [mlx5_core]
RIP: 0010:ioread32be+0x30/0x40
Code: 00 77 27 48 81 ff 00 00 01 00 76 07 0f b7 d7 ed 0f c8 c3 55 48 c7 c6 3b ee d5 9f 48 89 e5 e8 67 fc ff ff b8 ff ff ff ff 5d c3 <8b> 07 0f c8 c3 66 66 2e 0f 1f 84 00 00 00 00 00 48 81 fe ff ff 03
RSP: 0018:ffffb5b14c56fd78 EFLAGS: 00010292
RAX: ffffb5b141c18000 RBX: ffff8e9f78a801c0 RCX: 0000000000000000
RDX: 0000000000000001 RSI: ffff8e9f7ecd7628 RDI: ffffb5b141c18014
RBP: ffffb5b14c56fd90 R08: 0000000000000001 R09: 0000000000000000
R10: ffff8e9f372a2c30 R11: ffff8e9f87f4bc40 R12: ffff8e9f372a1fc0
R13: ffff8e9f78a80000 R14: ffffffffc07136a0 R15: ffff8e9f78ae6f20
FS:  0000000000000000(0000) GS:ffff8e9f7ecc0000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffffb5b141c18014 CR3: 00000001c8f82006 CR4: 00000000003606e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 ? mlx5_health_try_recover+0x4d/0x270 [mlx5_core]
 mlx5_fw_fatal_reporter_recover+0x16/0x20 [mlx5_core]
 devlink_health_reporter_recover+0x1c/0x50
 devlink_health_report+0xfb/0x240
 mlx5_fw_fatal_reporter_err_work+0x65/0xd0 [mlx5_core]
 process_one_work+0x1fb/0x4e0
 ? process_one_work+0x16b/0x4e0
 worker_thread+0x4f/0x3d0
 kthread+0x10d/0x140
 ? process_one_work+0x4e0/0x4e0
 ? kthread_cancel_delayed_work_sync+0x20/0x20
 ret_from_fork+0x1f/0x30
Modules linked in: nfsv3 rpcsec_gss_krb5 nfsv4 nfs fscache 8021q garp mrp stp llc ipmi_devintf ipmi_msghandler rpcrdma rdma_ucm ib_iser rdma_cm ib_umad iw_cm ib_ipoib libiscsi scsi_transport_iscsi ib_cm mlx5_ib ib_uverbs ib_core mlx5_core sb_edac crct10dif_pclmul crc32_pclmul ghash_clmulni_intel aesni_intel aes_x86_64 mlxfw crypto_simd cryptd glue_helper input_leds hyperv_fb intel_rapl_perf joydev serio_raw pci_hyperv pci_hyperv_mini mac_hid hv_balloon nfsd auth_rpcgss nfs_acl lockd grace sunrpc sch_fq_codel ip_tables x_tables autofs4 hv_utils hid_generic hv_storvsc ptp hid_hyperv hid hv_netvsc hyperv_keyboard pps_core scsi_transport_fc psmouse hv_vmbus i2c_piix4 floppy pata_acpi
CR2: ffffb5b141c18014
---[ end trace b12c5503157cad24 ]---
RIP: 0010:ioread32be+0x30/0x40
Code: 00 77 27 48 81 ff 00 00 01 00 76 07 0f b7 d7 ed 0f c8 c3 55 48 c7 c6 3b ee d5 9f 48 89 e5 e8 67 fc ff ff b8 ff ff ff ff 5d c3 <8b> 07 0f c8 c3 66 66 2e 0f 1f 84 00 00 00 00 00 48 81 fe ff ff 03
RSP: 0018:ffffb5b14c56fd78 EFLAGS: 00010292
RAX: ffffb5b141c18000 RBX: ffff8e9f78a801c0 RCX: 0000000000000000
RDX: 0000000000000001 RSI: ffff8e9f7ecd7628 RDI: ffffb5b141c18014
RBP: ffffb5b14c56fd90 R08: 0000000000000001 R09: 0000000000000000
R10: ffff8e9f372a2c30 R11: ffff8e9f87f4bc40 R12: ffff8e9f372a1fc0
R13: ffff8e9f78a80000 R14: ffffffffc07136a0 R15: ffff8e9f78ae6f20
FS:  0000000000000000(0000) GS:ffff8e9f7ecc0000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffffb5b141c18014 CR3: 00000001c8f82006 CR4: 00000000003606e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
BUG: sleeping function called from invalid context at ./include/linux/percpu-rwsem.h:38
in_atomic(): 0, irqs_disabled(): 1, pid: 6755, name: kworker/u128:2
INFO: lockdep is turned off.
CPU: 3 PID: 6755 Comm: kworker/u128:2 Tainted: G      D           5.2.0-net-next-mlx5-hv_stats-over-last-worked-hyperv #1
Hardware name: Microsoft Corporation Virtual Machine/Virtual Machine, BIOS 090006  04/28/2016
Workqueue: mlx5_healtha050:00:02.0 mlx5_fw_fatal_reporter_err_work [mlx5_core]
Call Trace:
 dump_stack+0x63/0x88
 ___might_sleep+0x10a/0x130
 __might_sleep+0x4a/0x80
 exit_signals+0x33/0x230
 ? blocking_notifier_call_chain+0x16/0x20
 do_exit+0xb1/0xc30
 ? kthread+0x10d/0x140
 ? process_one_work+0x4e0/0x4e0

Fixes: 52c368dc3da7 ("net/mlx5: Move health and page alloc init to mdev_init")
Signed-off-by: Shay Drory <shayd@mellanox.com>
Reviewed-by: Moshe Shemesh <moshe@mellanox.com>
Signed-off-by: Saeed Mahameed <saeedm@mellanox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoselftests/net: in rxtimestamp getopt_long needs terminating null entry
tannerlove [Tue, 9 Jun 2020 21:21:32 +0000 (17:21 -0400)]
selftests/net: in rxtimestamp getopt_long needs terminating null entry

[ Upstream commit 865a6cbb2288f8af7f9dc3b153c61b7014fdcf1e ]

getopt_long requires the last element to be filled with zeros.
Otherwise, passing an unrecognized option can cause a segfault.

Fixes: 16e781224198 ("selftests/net: Add a test to validate behavior of rx timestamps")
Signed-off-by: Tanner Love <tannerlove@google.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agodccp: Fix possible memleak in dccp_init and dccp_fini
Wang Hai [Tue, 9 Jun 2020 14:18:16 +0000 (22:18 +0800)]
dccp: Fix possible memleak in dccp_init and dccp_fini

[ Upstream commit c96b6acc8f89a4a7f6258dfe1d077654c11415be ]

There are some memory leaks in dccp_init() and dccp_fini().

In dccp_fini() and the error handling path in dccp_init(), free lhash2
is missing. Add inet_hashinfo2_free_mod() to do it.

If inet_hashinfo2_init_mod() failed in dccp_init(),
percpu_counter_destroy() should be called to destroy dccp_orphan_count.
It need to goto out_free_percpu when inet_hashinfo2_init_mod() failed.

Fixes: c92c81df93df ("net: dccp: fix kernel crash on module load")
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Wang Hai <wanghai38@huawei.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agocrypto: virtio: Fix dest length calculation in __virtio_crypto_skcipher_do_req()
Longpeng(Mike) [Tue, 2 Jun 2020 07:05:01 +0000 (15:05 +0800)]
crypto: virtio: Fix dest length calculation in __virtio_crypto_skcipher_do_req()

[ Upstream commit d90ca42012db2863a9a30b564a2ace6016594bda ]

The src/dst length is not aligned with AES_BLOCK_SIZE(which is 16) in some
testcases in tcrypto.ko.

For example, the src/dst length of one of cts(cbc(aes))'s testcase is 17, the
crypto_virtio driver will set @src_data_len=16 but @dst_data_len=17 in this
case and get a wrong at then end.

  SRC: pp pp pp pp pp pp pp pp pp pp pp pp pp pp pp pp pp (17 bytes)
  EXP: cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc pp (17 bytes)
  DST: cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc cc 00 (pollute the last bytes)
  (pp: plaintext  cc:ciphertext)

Fix this issue by limit the length of dest buffer.

Fixes: dbaf0624ffa5 ("crypto: add virtio-crypto driver")
Cc: Gonglei <arei.gonglei@huawei.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: virtualization@lists.linux-foundation.org
Cc: linux-kernel@vger.kernel.org
Cc: stable@vger.kernel.org
Signed-off-by: Longpeng(Mike) <longpeng2@huawei.com>
Link: https://lore.kernel.org/r/20200602070501.2023-4-longpeng2@huawei.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agocrypto: virtio: Fix src/dst scatterlist calculation in __virtio_crypto_skcipher_do_req()
Longpeng(Mike) [Tue, 2 Jun 2020 07:04:59 +0000 (15:04 +0800)]
crypto: virtio: Fix src/dst scatterlist calculation in __virtio_crypto_skcipher_do_req()

[ Upstream commit b02989f37fc5e865ceeee9070907e4493b3a21e2 ]

The system will crash when the users insmod crypto/tcrypt.ko with mode=38
( testing "cts(cbc(aes))" ).

Usually the next entry of one sg will be @sg@ + 1, but if this sg element
is part of a chained scatterlist, it could jump to the start of a new
scatterlist array. Fix it by sg_next() on calculation of src/dst
scatterlist.

Fixes: dbaf0624ffa5 ("crypto: add virtio-crypto driver")
Reported-by: LABBE Corentin <clabbe@baylibre.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: virtualization@lists.linux-foundation.org
Cc: linux-kernel@vger.kernel.org
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200123101000.GB24255@Red
Signed-off-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Longpeng(Mike) <longpeng2@huawei.com>
Link: https://lore.kernel.org/r/20200602070501.2023-2-longpeng2@huawei.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agocrypto: virtio: Fix use-after-free in virtio_crypto_skcipher_finalize_req()
Longpeng(Mike) [Tue, 2 Jun 2020 07:05:00 +0000 (15:05 +0800)]
crypto: virtio: Fix use-after-free in virtio_crypto_skcipher_finalize_req()

[ Upstream commit 8c855f0720ff006d75d0a2512c7f6c4f60ff60ee ]

The system'll crash when the users insmod crypto/tcrypto.ko with mode=155
( testing "authenc(hmac(sha1),cbc(aes))" ). It's caused by reuse the memory
of request structure.

In crypto_authenc_init_tfm(), the reqsize is set to:
  [PART 1] sizeof(authenc_request_ctx) +
  [PART 2] ictx->reqoff +
  [PART 3] MAX(ahash part, skcipher part)
and the 'PART 3' is used by both ahash and skcipher in turn.

When the virtio_crypto driver finish skcipher req, it'll call ->complete
callback(in crypto_finalize_skcipher_request) and then free its
resources whose pointers are recorded in 'skcipher parts'.

However, the ->complete is 'crypto_authenc_encrypt_done' in this case,
it will use the 'ahash part' of the request and change its content,
so virtio_crypto driver will get the wrong pointer after ->complete
finish and mistakenly free some other's memory. So the system will crash
when these memory will be used again.

The resources which need to be cleaned up are not used any more. But the
pointers of these resources may be changed in the function
"crypto_finalize_skcipher_request". Thus release specific resources before
calling this function.

Fixes: dbaf0624ffa5 ("crypto: add virtio-crypto driver")
Reported-by: LABBE Corentin <clabbe@baylibre.com>
Cc: Gonglei <arei.gonglei@huawei.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: "Michael S. Tsirkin" <mst@redhat.com>
Cc: Jason Wang <jasowang@redhat.com>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: virtualization@lists.linux-foundation.org
Cc: linux-kernel@vger.kernel.org
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200123101000.GB24255@Red
Acked-by: Gonglei <arei.gonglei@huawei.com>
Signed-off-by: Longpeng(Mike) <longpeng2@huawei.com>
Link: https://lore.kernel.org/r/20200602070501.2023-3-longpeng2@huawei.com
Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agofirmware: imx: scu: Fix corruption of header
Franck LENORMAND [Wed, 25 Mar 2020 22:00:05 +0000 (00:00 +0200)]
firmware: imx: scu: Fix corruption of header

[ Upstream commit f5f27b79eab80de0287c243a22169e4876b08d5e ]

The header of the message to send can be changed if the
response is longer than the request:
 - 1st word, the header is sent
 - the remaining words of the message are sent
 - the response is received asynchronously during the
   execution of the loop, changing the size field in
   the header
 - the for loop test the termination condition using
   the corrupted header

It is the case for the API build_info which has just a
header as request but 3 words in response.

This issue is fixed storing the header locally instead of
using a pointer on it.

Fixes: edbee095fafb (firmware: imx: add SCU firmware driver support)

Signed-off-by: Franck LENORMAND <franck.lenormand@nxp.com>
Reviewed-by: Leonard Crestez <leonard.crestez@nxp.com>
Signed-off-by: Leonard Crestez <leonard.crestez@nxp.com>
Cc: stable@vger.kernel.org
Reviewed-by: Dong Aisheng <aisheng.dong@nxp.com>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agofirmware: imx-scu: Support one TX and one RX
Peng Fan [Thu, 19 Mar 2020 07:49:53 +0000 (15:49 +0800)]
firmware: imx-scu: Support one TX and one RX

[ Upstream commit f25a066d1a07affb7bea4e5d9c179c3338338e23 ]

Current imx-scu requires four TX and four RX to communicate with
SCU. This is low efficient and causes lots of mailbox interrupts.

With imx-mailbox driver could support one TX to use all four transmit
registers and one RX to use all four receive registers, imx-scu
could use one TX and one RX.

Signed-off-by: Peng Fan <peng.fan@nxp.com>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agofirmware: imx: warn on unexpected RX
Leonard Crestez [Wed, 4 Sep 2019 07:54:58 +0000 (10:54 +0300)]
firmware: imx: warn on unexpected RX

[ Upstream commit cf0fd404455ce13850cc15423a3c2958933de384 ]

The imx_scu_call_rpc function returns the result inside the
same "msg" struct containing the transmitted message. This is
implemented by holding a pointer to msg (which is usually on the stack)
in sc_imx_rpc and writing to it from imx_scu_rx_callback.

This means that if the have_resp parameter is incorrect or SCU sends an
unexpected response for any reason the most likely result is kernel stack
corruption.

Fix this by only setting sc_imx_rpc.msg for the duration of the
imx_scu_call_rpc call and warning in imx_scu_rx_callback if unset.

Print the unexpected response data to help debugging.

Signed-off-by: Leonard Crestez <leonard.crestez@nxp.com>
Acked-by: Anson Huang <Anson.Huang@nxp.com>
Signed-off-by: Shawn Guo <shawnguo@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agox86/{mce,mm}: Unmap the entire page if the whole page is affected and poisoned
Tony Luck [Wed, 20 May 2020 16:35:46 +0000 (09:35 -0700)]
x86/{mce,mm}: Unmap the entire page if the whole page is affected and poisoned

commit 17fae1294ad9d711b2c3dd0edef479d40c76a5e8 upstream.

An interesting thing happened when a guest Linux instance took a machine
check. The VMM unmapped the bad page from guest physical space and
passed the machine check to the guest.

Linux took all the normal actions to offline the page from the process
that was using it. But then guest Linux crashed because it said there
was a second machine check inside the kernel with this stack trace:

do_memory_failure
    set_mce_nospec
         set_memory_uc
              _set_memory_uc
                   change_page_attr_set_clr
                        cpa_flush
                             clflush_cache_range_opt

This was odd, because a CLFLUSH instruction shouldn't raise a machine
check (it isn't consuming the data). Further investigation showed that
the VMM had passed in another machine check because is appeared that the
guest was accessing the bad page.

Fix is to check the scope of the poison by checking the MCi_MISC register.
If the entire page is affected, then unmap the page. If only part of the
page is affected, then mark the page as uncacheable.

This assumes that VMMs will do the logical thing and pass in the "whole
page scope" via the MCi_MISC register (since they unmapped the entire
page).

  [ bp: Adjust to x86/entry changes. ]

Fixes: 284ce4011ba6 ("x86/memory_failure: Introduce {set, clear}_mce_nospec()")
Reported-by: Jue Wang <juew@google.com>
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Jue Wang <juew@google.com>
Cc: <stable@vger.kernel.org>
Link: https://lkml.kernel.org/r/20200520163546.GA7977@agluck-desk2.amr.corp.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agocrypto: drbg - fix error return code in drbg_alloc_state()
Wei Yongjun [Thu, 30 Apr 2020 08:13:53 +0000 (08:13 +0000)]
crypto: drbg - fix error return code in drbg_alloc_state()

commit e0664ebcea6ac5e16da703409fb4bd61f8cd37d9 upstream.

Fix to return negative error code -ENOMEM from the kzalloc error handling
case instead of 0, as done elsewhere in this function.

Reported-by: Xiumei Mu <xmu@redhat.com>
Fixes: db07cd26ac6a ("crypto: drbg - add FIPS 140-2 CTRNG for noise source")
Cc: <stable@vger.kernel.org>
Signed-off-by: Wei Yongjun <weiyongjun1@huawei.com>
Reviewed-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agocrypto: algapi - Avoid spurious modprobe on LOADED
Eric Biggers [Tue, 7 Apr 2020 06:02:40 +0000 (23:02 -0700)]
crypto: algapi - Avoid spurious modprobe on LOADED

commit beeb460cd12ac9b91640b484b6a52dcba9d9fc8f upstream.

Currently after any algorithm is registered and tested, there's an
unnecessary request_module("cryptomgr") even if it's already loaded.
Also, CRYPTO_MSG_ALG_LOADED is sent twice, and thus if the algorithm is
"crct10dif", lib/crc-t10dif.c replaces the tfm twice rather than once.

This occurs because CRYPTO_MSG_ALG_LOADED is sent using
crypto_probing_notify(), which tries to load "cryptomgr" if the
notification is not handled (NOTIFY_DONE).  This doesn't make sense
because "cryptomgr" doesn't handle this notification.

Fix this by using crypto_notify() instead of crypto_probing_notify().

Fixes: dd8b083f9a5e ("crypto: api - Introduce notifier for new crypto algorithms")
Cc: <stable@vger.kernel.org> # v4.20+
Cc: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agocrypto: cavium/nitrox - Fix 'nitrox_get_first_device()' when ndevlist is fully iterated
Christophe JAILLET [Sat, 30 May 2020 13:35:37 +0000 (15:35 +0200)]
crypto: cavium/nitrox - Fix 'nitrox_get_first_device()' when ndevlist is fully iterated

commit 320bdbd816156f9ca07e5fed7bfb449f2908dda7 upstream.

When a list is completely iterated with 'list_for_each_entry(x, ...)', x is
not NULL at the end.

While at it, remove a useless initialization of the ndev variable. It
is overridden by 'list_for_each_entry'.

Fixes: f2663872f073 ("crypto: cavium - Register the CNN55XX supported crypto algorithms.")
Cc: <stable@vger.kernel.org>
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agogup: document and work around "COW can break either way" issue
Linus Torvalds [Thu, 28 May 2020 01:29:34 +0000 (18:29 -0700)]
gup: document and work around "COW can break either way" issue

commit 17839856fd588f4ab6b789f482ed3ffd7c403e1f upstream.

Doing a "get_user_pages()" on a copy-on-write page for reading can be
ambiguous: the page can be COW'ed at any time afterwards, and the
direction of a COW event isn't defined.

Yes, whoever writes to it will generally do the COW, but if the thread
that did the get_user_pages() unmapped the page before the write (and
that could happen due to memory pressure in addition to any outright
action), the writer could also just take over the old page instead.

End result: the get_user_pages() call might result in a page pointer
that is no longer associated with the original VM, and is associated
with - and controlled by - another VM having taken it over instead.

So when doing a get_user_pages() on a COW mapping, the only really safe
thing to do would be to break the COW when getting the page, even when
only getting it for reading.

At the same time, some users simply don't even care.

For example, the perf code wants to look up the page not because it
cares about the page, but because the code simply wants to look up the
physical address of the access for informational purposes, and doesn't
really care about races when a page might be unmapped and remapped
elsewhere.

This adds logic to force a COW event by setting FOLL_WRITE on any
copy-on-write mapping when FOLL_GET (or FOLL_PIN) is used to get a page
pointer as a result.

The current semantics end up being:

 - __get_user_pages_fast(): no change. If you don't ask for a write,
   you won't break COW. You'd better know what you're doing.

 - get_user_pages_fast(): the fast-case "look it up in the page tables
   without anything getting mmap_sem" now refuses to follow a read-only
   page, since it might need COW breaking.  Which happens in the slow
   path - the fast path doesn't know if the memory might be COW or not.

 - get_user_pages() (including the slow-path fallback for gup_fast()):
   for a COW mapping, turn on FOLL_WRITE for FOLL_GET/FOLL_PIN, with
   very similar semantics to FOLL_FORCE.

If it turns out that we want finer granularity (ie "only break COW when
it might actually matter" - things like the zero page are special and
don't need to be broken) we might need to push these semantics deeper
into the lookup fault path.  So if people care enough, it's possible
that we might end up adding a new internal FOLL_BREAK_COW flag to go
with the internal FOLL_COW flag we already have for tracking "I had a
COW".

Alternatively, if it turns out that different callers might want to
explicitly control the forced COW break behavior, we might even want to
make such a flag visible to the users of get_user_pages() instead of
using the above default semantics.

But for now, this is mostly commentary on the issue (this commit message
being a lot bigger than the patch, and that patch in turn is almost all
comments), with that minimal "enable COW breaking early" logic using the
existing FOLL_WRITE behavior.

[ It might be worth noting that we've always had this ambiguity, and it
  could arguably be seen as a user-space issue.

  You only get private COW mappings that could break either way in
  situations where user space is doing cooperative things (ie fork()
  before an execve() etc), but it _is_ surprising and very subtle, and
  fork() is supposed to give you independent address spaces.

  So let's treat this as a kernel issue and make the semantics of
  get_user_pages() easier to understand. Note that obviously a true
  shared mapping will still get a page that can change under us, so this
  does _not_ mean that get_user_pages() somehow returns any "stable"
  page ]

Reported-by: Jann Horn <jannh@google.com>
Tested-by: Christoph Hellwig <hch@lst.de>
Acked-by: Oleg Nesterov <oleg@redhat.com>
Acked-by: Kirill Shutemov <kirill@shutemov.name>
Acked-by: Jan Kara <jack@suse.cz>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Matthew Wilcox <willy@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoPM: runtime: clk: Fix clk_pm_runtime_get() error path
Rafael J. Wysocki [Thu, 21 May 2020 17:08:09 +0000 (19:08 +0200)]
PM: runtime: clk: Fix clk_pm_runtime_get() error path

commit 64c7d7ea22d86cacb65d0c097cc447bc0e6d8abd upstream.

clk_pm_runtime_get() assumes that the PM-runtime usage counter will
be dropped by pm_runtime_get_sync() on errors, which is not the case,
so PM-runtime references to devices acquired by the former are leaked
on errors returned by the latter.

Fix this by modifying clk_pm_runtime_get() to drop the reference if
pm_runtime_get_sync() returns an error.

Fixes: 9a34b45397e5 clk: Add support for runtime PM
Cc: 4.15+ <stable@vger.kernel.org> # 4.15+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agospi: bcm-qspi: when tx/rx buffer is NULL set to 0
Justin Chen [Mon, 20 Apr 2020 19:08:49 +0000 (15:08 -0400)]
spi: bcm-qspi: when tx/rx buffer is NULL set to 0

commit 4df3bea7f9d2ddd9ac2c29ba945c7c4db2def29c upstream.

Currently we set the tx/rx buffer to 0xff when NULL. This causes
problems with some spi slaves where 0xff is a valid command. Looking
at other drivers, the tx/rx buffer is usually set to 0x00 when NULL.
Following this convention solves the issue.

Fixes: fa236a7ef240 ("spi: bcm-qspi: Add Broadcom MSPI driver")
Signed-off-by: Justin Chen <justinpopo6@gmail.com>
Signed-off-by: Kamal Dasu <kdasu.kdev@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200420190853.45614-6-kdasu.kdev@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agospi: bcm-qspi: Handle clock probe deferral
Florian Fainelli [Mon, 20 Apr 2020 19:08:45 +0000 (15:08 -0400)]
spi: bcm-qspi: Handle clock probe deferral

commit 0392727c261bab65a35cd4f82ee9459bc237591d upstream.

The clock provider may not be ready by the time spi-bcm-qspi gets
probed, handle probe deferral using devm_clk_get_optional().

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Kamal Dasu <kdasu.kdev@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200420190853.45614-2-kdasu.kdev@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agospi: bcm2835aux: Fix controller unregister order
Lukas Wunner [Fri, 15 May 2020 15:58:03 +0000 (17:58 +0200)]
spi: bcm2835aux: Fix controller unregister order

commit b9dd3f6d417258ad0beeb292a1bc74200149f15d upstream.

The BCM2835aux SPI driver uses devm_spi_register_master() on bind.
As a consequence, on unbind, __device_release_driver() first invokes
bcm2835aux_spi_remove() before unregistering the SPI controller via
devres_release_all().

This order is incorrect:  bcm2835aux_spi_remove() turns off the SPI
controller, including its interrupts and clock.  The SPI controller
is thus no longer usable.

When the SPI controller is subsequently unregistered, it unbinds all
its slave devices.  If their drivers need to access the SPI bus,
e.g. to quiesce their interrupts, unbinding will fail.

As a rule, devm_spi_register_master() must not be used if the
->remove() hook performs teardown steps which shall be performed
after unbinding of slaves.

Fix by using the non-devm variant spi_register_master().  Note that the
struct spi_master as well as the driver-private data are not freed until
after bcm2835aux_spi_remove() has finished, so accessing them is safe.

Fixes: 1ea29b39f4c8 ("spi: bcm2835aux: add bcm2835 auxiliary spi device driver")
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: stable@vger.kernel.org # v4.4+
Cc: Martin Sperl <kernel@martin.sperl.org>
Link: https://lore.kernel.org/r/32f27f4d8242e4d75f9a53f7e8f1f77483b08669.1589557526.git.lukas@wunner.de
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agospi: bcm2835: Fix controller unregister order
Lukas Wunner [Fri, 15 May 2020 15:58:02 +0000 (17:58 +0200)]
spi: bcm2835: Fix controller unregister order

commit 9dd277ff92d06f6aa95b39936ad83981d781f49b upstream.

The BCM2835 SPI driver uses devm_spi_register_controller() on bind.
As a consequence, on unbind, __device_release_driver() first invokes
bcm2835_spi_remove() before unregistering the SPI controller via
devres_release_all().

This order is incorrect:  bcm2835_spi_remove() tears down the DMA
channels and turns off the SPI controller, including its interrupts
and clock.  The SPI controller is thus no longer usable.

When the SPI controller is subsequently unregistered, it unbinds all
its slave devices.  If their drivers need to access the SPI bus,
e.g. to quiesce their interrupts, unbinding will fail.

As a rule, devm_spi_register_controller() must not be used if the
->remove() hook performs teardown steps which shall be performed
after unbinding of slaves.

Fix by using the non-devm variant spi_register_controller().  Note that
the struct spi_controller as well as the driver-private data are not
freed until after bcm2835_spi_remove() has finished, so accessing them
is safe.

Fixes: 247263dba208 ("spi: bcm2835: use devm_spi_register_master()")
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: stable@vger.kernel.org # v3.13+
Link: https://lore.kernel.org/r/2397dd70cdbe95e0bc4da2b9fca0f31cb94e5aed.1589557526.git.lukas@wunner.de
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agospi: pxa2xx: Fix runtime PM ref imbalance on probe error
Lukas Wunner [Mon, 25 May 2020 12:25:03 +0000 (14:25 +0200)]
spi: pxa2xx: Fix runtime PM ref imbalance on probe error

commit 65e318e17358a3fd4fcb5a69d89b14016dee2f06 upstream.

The PXA2xx SPI driver releases a runtime PM ref in the probe error path
even though it hasn't acquired a ref earlier.

Apparently commit e2b714afee32 ("spi: pxa2xx: Disable runtime PM if
controller registration fails") sought to copy-paste the invocation of
pm_runtime_disable() from pxa2xx_spi_remove(), but erroneously copied
the call to pm_runtime_put_noidle() as well.  Drop it.

Fixes: e2b714afee32 ("spi: pxa2xx: Disable runtime PM if controller registration fails")
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Reviewed-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: stable@vger.kernel.org # v4.17+
Cc: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Link: https://lore.kernel.org/r/58b2ac6942ca1f91aaeeafe512144bc5343e1d84.1590408496.git.lukas@wunner.de
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agospi: pxa2xx: Fix controller unregister order
Lukas Wunner [Mon, 25 May 2020 12:25:02 +0000 (14:25 +0200)]
spi: pxa2xx: Fix controller unregister order

commit 32e5b57232c0411e7dea96625c415510430ac079 upstream.

The PXA2xx SPI driver uses devm_spi_register_controller() on bind.
As a consequence, on unbind, __device_release_driver() first invokes
pxa2xx_spi_remove() before unregistering the SPI controller via
devres_release_all().

This order is incorrect:  pxa2xx_spi_remove() disables the chip,
rendering the SPI bus inaccessible even though the SPI controller is
still registered.  When the SPI controller is subsequently unregistered,
it unbinds all its slave devices.  Because their drivers cannot access
the SPI bus, e.g. to quiesce interrupts, the slave devices may be left
in an improper state.

As a rule, devm_spi_register_controller() must not be used if the
->remove() hook performs teardown steps which shall be performed after
unregistering the controller and specifically after unbinding of slaves.

Fix by reverting to the non-devm variant of spi_register_controller().

An alternative approach would be to use device-managed functions for all
steps in pxa2xx_spi_remove(), e.g. by calling devm_add_action_or_reset()
on probe.  However that approach would add more LoC to the driver and
it wouldn't lend itself as well to backporting to stable.

The improper use of devm_spi_register_controller() was introduced in 2013
by commit a807fcd090d6 ("spi: pxa2xx: use devm_spi_register_master()"),
but all earlier versions of the driver going back to 2006 were likewise
broken because they invoked spi_unregister_master() at the end of
pxa2xx_spi_remove(), rather than at the beginning.

Fixes: e0c9905e87ac ("[PATCH] SPI: add PXA2xx SSP SPI Driver")
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: stable@vger.kernel.org # v2.6.17+
Cc: Tsuchiya Yuto <kitakar@gmail.com>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=206403#c1
Link: https://lore.kernel.org/r/834c446b1cf3284d2660f1bee1ebe3e737cd02a9.1590408496.git.lukas@wunner.de
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agospi: Fix controller unregister order
Lukas Wunner [Fri, 15 May 2020 15:58:01 +0000 (17:58 +0200)]
spi: Fix controller unregister order

commit 84855678add8aba927faf76bc2f130a40f94b6f7 upstream.

When an SPI controller unregisters, it unbinds all its slave devices.
For this, their drivers may need to access the SPI bus, e.g. to quiesce
interrupts.

However since commit ffbbdd21329f ("spi: create a message queueing
infrastructure"), spi_destroy_queue() is executed before unbinding the
slaves.  It sets ctlr->running = false, thereby preventing SPI bus
access and causing unbinding of slave devices to fail.

Fix by unbinding slaves before calling spi_destroy_queue().

Fixes: ffbbdd21329f ("spi: create a message queueing infrastructure")
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: stable@vger.kernel.org # v3.4+
Cc: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/8aaf9d44c153fe233b17bc2dec4eb679898d7e7b.1589557526.git.lukas@wunner.de
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agospi: dw: Fix controller unregister order
Lukas Wunner [Mon, 25 May 2020 12:25:01 +0000 (14:25 +0200)]
spi: dw: Fix controller unregister order

commit ca8b19d61e3fce5d2d7790cde27a0b57bcb3f341 upstream.

The Designware SPI driver uses devm_spi_register_controller() on bind.
As a consequence, on unbind, __device_release_driver() first invokes
dw_spi_remove_host() before unregistering the SPI controller via
devres_release_all().

This order is incorrect:  dw_spi_remove_host() shuts down the chip,
rendering the SPI bus inaccessible even though the SPI controller is
still registered.  When the SPI controller is subsequently unregistered,
it unbinds all its slave devices.  Because their drivers cannot access
the SPI bus, e.g. to quiesce interrupts, the slave devices may be left
in an improper state.

As a rule, devm_spi_register_controller() must not be used if the
->remove() hook performs teardown steps which shall be performed after
unregistering the controller and specifically after unbinding of slaves.

Fix by reverting to the non-devm variant of spi_register_controller().

An alternative approach would be to use device-managed functions for all
steps in dw_spi_remove_host(), e.g. by calling devm_add_action_or_reset()
on probe.  However that approach would add more LoC to the driver and
it wouldn't lend itself as well to backporting to stable.

Fixes: 04f421e7b0b1 ("spi: dw: use managed resources")
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: stable@vger.kernel.org # v3.14+
Cc: Baruch Siach <baruch@tkos.co.il>
Link: https://lore.kernel.org/r/3fff8cb8ae44a9893840d0688be15bb88c090a14.1590408496.git.lukas@wunner.de
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonilfs2: fix null pointer dereference at nilfs_segctor_do_construct()
Ryusuke Konishi [Thu, 11 Jun 2020 01:41:35 +0000 (18:41 -0700)]
nilfs2: fix null pointer dereference at nilfs_segctor_do_construct()

commit 8301c719a2bd131436438e49130ee381d30933f5 upstream.

After commit c3aab9a0bd91 ("mm/filemap.c: don't initiate writeback if
mapping has no dirty pages"), the following null pointer dereference has
been reported on nilfs2:

  BUG: kernel NULL pointer dereference, address: 00000000000000a8
  #PF: supervisor read access in kernel mode
  #PF: error_code(0x0000) - not-present page
  PGD 0 P4D 0
  Oops: 0000 [#1] SMP PTI
  ...
  RIP: 0010:percpu_counter_add_batch+0xa/0x60
  ...
  Call Trace:
    __test_set_page_writeback+0x2d3/0x330
    nilfs_segctor_do_construct+0x10d3/0x2110 [nilfs2]
    nilfs_segctor_construct+0x168/0x260 [nilfs2]
    nilfs_segctor_thread+0x127/0x3b0 [nilfs2]
    kthread+0xf8/0x130
    ...

This crash turned out to be caused by set_page_writeback() call for
segment summary buffers at nilfs_segctor_prepare_write().

set_page_writeback() can call inc_wb_stat(inode_to_wb(inode),
WB_WRITEBACK) where inode_to_wb(inode) is NULL if the inode of
underlying block device does not have an associated wb.

This fixes the issue by calling inode_attach_wb() in advance to ensure
to associate the bdev inode with its wb.

Fixes: c3aab9a0bd91 ("mm/filemap.c: don't initiate writeback if mapping has no dirty pages")
Reported-by: Walton Hoops <me@waltonhoops.com>
Reported-by: Tomas Hlavaty <tom@logand.com>
Reported-by: ARAI Shun-ichi <hermes@ceres.dti.ne.jp>
Reported-by: Hideki EIRAKU <hdk1983@gmail.com>
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Tested-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Cc: <stable@vger.kernel.org> [5.4+]
Link: http://lkml.kernel.org/r/20200608.011819.1399059588922299158.konishi.ryusuke@gmail.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agolib/lzo: fix ambiguous encoding bug in lzo-rle
Dave Rodgman [Fri, 12 Jun 2020 00:34:54 +0000 (17:34 -0700)]
lib/lzo: fix ambiguous encoding bug in lzo-rle

commit b5265c813ce4efbfa2e46fd27cdf9a7f44a35d2e upstream.

In some rare cases, for input data over 32 KB, lzo-rle could encode two
different inputs to the same compressed representation, so that
decompression is then ambiguous (i.e.  data may be corrupted - although
zram is not affected because it operates over 4 KB pages).

This modifies the compressor without changing the decompressor or the
bitstream format, such that:

 - there is no change to how data produced by the old compressor is
   decompressed

 - an old decompressor will correctly decode data from the updated
   compressor

 - performance and compression ratio are not affected

 - we avoid introducing a new bitstream format

In testing over 12.8M real-world files totalling 903 GB, three files
were affected by this bug.  I also constructed 37M semi-random 64 KB
files totalling 2.27 TB, and saw no affected files.  Finally I tested
over files constructed to contain each of the ~1024 possible bad input
sequences; for all of these cases, updated lzo-rle worked correctly.

There is no significant impact to performance or compression ratio.

Signed-off-by: Dave Rodgman <dave.rodgman@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Dave Rodgman <dave.rodgman@arm.com>
Cc: Willy Tarreau <w@1wt.eu>
Cc: Sergey Senozhatsky <sergey.senozhatsky.work@gmail.com>
Cc: Markus F.X.J. Oberhumer <markus@oberhumer.com>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Cc: Chao Yu <yuchao0@huawei.com>
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20200507100203.29785-1-dave.rodgman@arm.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoarm64: acpi: fix UBSAN warning
Nick Desaulniers [Mon, 8 Jun 2020 20:38:17 +0000 (13:38 -0700)]
arm64: acpi: fix UBSAN warning

commit a194c33f45f83068ef13bf1d16e26d4ca3ecc098 upstream.

Will reported a UBSAN warning:

UBSAN: null-ptr-deref in arch/arm64/kernel/smp.c:596:6
member access within null pointer of type 'struct acpi_madt_generic_interrupt'
CPU: 0 PID: 0 Comm: swapper Not tainted 5.7.0-rc6-00124-g96bc42ff0a82 #1
Call trace:
 dump_backtrace+0x0/0x384
 show_stack+0x28/0x38
 dump_stack+0xec/0x174
 handle_null_ptr_deref+0x134/0x174
 __ubsan_handle_type_mismatch_v1+0x84/0xa4
 acpi_parse_gic_cpu_interface+0x60/0xe8
 acpi_parse_entries_array+0x288/0x498
 acpi_table_parse_entries_array+0x178/0x1b4
 acpi_table_parse_madt+0xa4/0x110
 acpi_parse_and_init_cpus+0x38/0x100
 smp_init_cpus+0x74/0x258
 setup_arch+0x350/0x3ec
 start_kernel+0x98/0x6f4

This is from the use of the ACPI_OFFSET in
arch/arm64/include/asm/acpi.h. Replace its use with offsetof from
include/linux/stddef.h which should implement the same logic using
__builtin_offsetof, so that UBSAN wont warn.

Reported-by: Will Deacon <will@kernel.org>
Suggested-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Jeremy Linton <jeremy.linton@arm.com>
Acked-by: Lorenzo Pieralisi <lorenzo.pieralisi@arm.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/lkml/20200521100952.GA5360@willie-the-truck/
Link: https://lore.kernel.org/r/20200608203818.189423-1-ndesaulniers@google.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoACPI: PM: Avoid using power resources if there are none for D0
Rafael J. Wysocki [Thu, 4 Jun 2020 17:22:26 +0000 (19:22 +0200)]
ACPI: PM: Avoid using power resources if there are none for D0

commit 956ad9d98b73f59e442cc119c98ba1e04e94fe6d upstream.

As recently reported, some platforms provide a list of power
resources for device power state D3hot, through the _PR3 object,
but they do not provide a list of power resources for device power
state D0.

Among other things, this causes acpi_device_get_power() to return
D3hot as the current state of the device in question if all of the
D3hot power resources are "on", because it sees the power_resources
flag set and calls acpi_power_get_inferred_state() which finds that
D3hot is the shallowest power state with all of the associated power
resources turned "on", so that's what it returns.  Moreover, that
value takes precedence over the acpi_dev_pm_explicit_get() return
value, because it means a deeper power state.  The device may very
well be in D0 physically at that point, however.

Moreover, the presence of _PR3 without _PR0 for a given device
means that only one D3-level power state can be supported by it.
Namely, because there are no power resources to turn "off" when
transitioning the device from D0 into D3cold (which should be
supported since _PR3 is present), the evaluation of _PS3 should
be sufficient to put it straight into D3cold, but this means that
the effect of turning "on" the _PR3 power resources is unclear,
so it is better to avoid doing that altogether.  Consequently,
there is no practical way do distinguish D3cold from D3hot for
the device in question and the power states of it can be labeled
so that D3hot is the deepest supported one (and Linux assumes
that putting a device into D3hot via ACPI may cause power to be
removed from it anyway, for legacy reasons).

To work around the problem described above modify the ACPI
enumeration of devices so that power resources are only used
for device power management if the list of D0 power resources
is not empty and make it mart D3cold as supported only if that
is the case and the D3hot list of power resources is not empty
too.

Fixes: ef85bdbec444 ("ACPI / scan: Consolidate extraction of power resources lists")
Link: https://bugzilla.kernel.org/show_bug.cgi?id=205057
Link: https://lore.kernel.org/linux-acpi/20200603194659.185757-1-hdegoede@redhat.com/
Reported-by: Hans de Goede <hdegoede@redhat.com>
Tested-by: Hans de Goede <hdegoede@redhat.com>
Tested-by: youling257@gmail.com
Cc: 3.10+ <stable@vger.kernel.org> # 3.10+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoACPI: GED: add support for _Exx / _Lxx handler methods
Ard Biesheuvel [Fri, 15 May 2020 09:36:13 +0000 (11:36 +0200)]
ACPI: GED: add support for _Exx / _Lxx handler methods

commit ea6f3af4c5e63f6981c0b0ab8ebec438e2d5ef40 upstream.

Per the ACPI spec, interrupts in the range [0, 255] may be handled
in AML using individual methods whose naming is based on the format
_Exx or _Lxx, where xx is the hex representation of the interrupt
index.

Add support for this missing feature to our ACPI GED driver.

Cc: v4.9+ <stable@vger.kernel.org> # v4.9+
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoACPI: CPPC: Fix reference count leak in acpi_cppc_processor_probe()
Qiushi Wu [Wed, 27 May 2020 22:35:51 +0000 (17:35 -0500)]
ACPI: CPPC: Fix reference count leak in acpi_cppc_processor_probe()

commit 4d8be4bc94f74bb7d096e1c2e44457b530d5a170 upstream.

kobject_init_and_add() takes reference even when it fails.
If this function returns an error, kobject_put() must be called to
properly clean up the memory associated with the object. Previous
commit "b8eb718348b8" fixed a similar problem.

Fixes: 158c998ea44b ("ACPI / CPPC: add sysfs support to compute delivered performance")
Signed-off-by: Qiushi Wu <wu000273@umn.edu>
Cc: 4.10+ <stable@vger.kernel.org> # 4.10+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoACPI: sysfs: Fix reference count leak in acpi_sysfs_add_hotplug_profile()
Qiushi Wu [Wed, 27 May 2020 21:17:17 +0000 (16:17 -0500)]
ACPI: sysfs: Fix reference count leak in acpi_sysfs_add_hotplug_profile()

commit 6e6c25283dff866308c87b49434c7dbad4774cc0 upstream.

kobject_init_and_add() takes reference even when it fails.
Thus, when kobject_init_and_add() returns an error,
kobject_put() must be called to properly clean up the kobject.

Fixes: 3f8055c35836 ("ACPI / hotplug: Introduce user space interface for hotplug profiles")
Signed-off-by: Qiushi Wu <wu000273@umn.edu>
Cc: 3.10+ <stable@vger.kernel.org> # 3.10+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: usb-audio: Add vendor, product and profile name for HP Thunderbolt Dock
Kai-Heng Feng [Mon, 8 Jun 2020 06:26:28 +0000 (14:26 +0800)]
ALSA: usb-audio: Add vendor, product and profile name for HP Thunderbolt Dock

commit 0c5086f5699906ec8e31ea6509239489f060f2dc upstream.

The HP Thunderbolt Dock has two separate USB devices, one is for speaker
and one is for headset. Add names for them so userspace can apply UCM
settings.

Signed-off-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20200608062630.10806-1-kai.heng.feng@canonical.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: usb-audio: Fix inconsistent card PM state after resume
Takashi Iwai [Wed, 3 Jun 2020 15:37:08 +0000 (17:37 +0200)]
ALSA: usb-audio: Fix inconsistent card PM state after resume

commit 862b2509d157c629dd26d7ac6c6cdbf043d332eb upstream.

When a USB-audio interface gets runtime-suspended via auto-pm feature,
the driver suspends all functionality and increment
chip->num_suspended_intf.  Later on, when the system gets suspended to
S3, the driver increments chip->num_suspended_intf again, skips the
device changes, and sets the card power state to
SNDRV_CTL_POWER_D3hot.  In return, when the system gets resumed from
S3, the resume callback decrements chip->num_suspended_intf.  Since
this refcount is still not zero (it's been runtime-suspended), the
whole resume is skipped.  But there is a small pitfall here.

The problem is that the driver doesn't restore the card power state
after this resume call, leaving it as SNDRV_CTL_POWER_D3hot.  So,
even after the system resume finishes, the card instance still appears
as if it were system-suspended, and this confuses many ioctl accesses
that are blocked unexpectedly.

In details, we have two issues behind the scene: one is that the card
power state is changed only when the refcount becomes zero, and
another is that the prior auto-suspend check is kept in a boolean
flag.  Although the latter problem is almost negligible since the
auto-pm feature is imposed only on the primary interface, but this can
be a potential problem on the devices with multiple interfaces.

This patch addresses those issues by the following:

- Replace chip->autosuspended boolean flag with chip->system_suspend
  counter

- At the first system-suspend, chip->num_suspended_intf is recorded to
  chip->system_suspend

- At system-resume, the card power state is restored when the
  chip->num_suspended_intf refcount reaches to chip->system_suspend,
  i.e. the state returns to the auto-suspended

Also, the patch fixes yet another hidden problem by the code
refactoring along with the fixes above: namely, when some resume
procedure failed, the driver left chip->num_suspended_intf that was
already decreased, and it might lead to the refcount unbalance.
In the new code, the refcount decrement is done after the whole resume
procedure, and the problem is avoided as well.

Fixes: 0662292aec05 ("ALSA: usb-audio: Handle normal and auto-suspend equally")
Reported-and-tested-by: Macpaul Lin <macpaul.lin@mediatek.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20200603153709.6293-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: pcm: fix snd_pcm_link() lockdep splat
Michał Mirosław [Mon, 8 Jun 2020 10:06:32 +0000 (12:06 +0200)]
ALSA: pcm: fix snd_pcm_link() lockdep splat

commit e18035cf5cb3d2bf8e4f4d350a23608bd208b934 upstream.

Add and use snd_pcm_stream_lock_nested() in snd_pcm_link/unlink
implementation.  The code is fine, but generates a lockdep complaint:

============================================
WARNING: possible recursive locking detected
5.7.1mq+ #381 Tainted: G           O
--------------------------------------------
pulseaudio/4180 is trying to acquire lock:
ffff888402d6f508 (&group->lock){-...}-{2:2}, at: snd_pcm_common_ioctl+0xda8/0xee0 [snd_pcm]

but task is already holding lock:
ffff8883f7a8cf18 (&group->lock){-...}-{2:2}, at: snd_pcm_common_ioctl+0xe4e/0xee0 [snd_pcm]

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(&group->lock);
  lock(&group->lock);

 *** DEADLOCK ***

 May be due to missing lock nesting notation

2 locks held by pulseaudio/4180:
 #0: ffffffffa1a05190 (snd_pcm_link_rwsem){++++}-{3:3}, at: snd_pcm_common_ioctl+0xca0/0xee0 [snd_pcm]
 #1: ffff8883f7a8cf18 (&group->lock){-...}-{2:2}, at: snd_pcm_common_ioctl+0xe4e/0xee0 [snd_pcm]
[...]

Cc: stable@vger.kernel.org
Fixes: f57f3df03a8e ("ALSA: pcm: More fine-grained PCM link locking")
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Link: https://lore.kernel.org/r/37252c65941e58473b1219ca9fab03d48f47e3e3.1591610330.git.mirq-linux@rere.qmqm.pl
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
4 years agoALSA: pcm: disallow linking stream to itself
Michał Mirosław [Mon, 8 Jun 2020 16:50:39 +0000 (18:50 +0200)]
ALSA: pcm: disallow linking stream to itself

commit 951e2736f4b11b58dc44d41964fa17c3527d882a upstream.

Prevent SNDRV_PCM_IOCTL_LINK linking stream to itself - the code
can't handle it. Fixed commit is not where bug was introduced, but
changes the context significantly.

Cc: stable@vger.kernel.org
Fixes: 0888c321de70 ("pcm_native: switch to fdget()/fdput()")
Signed-off-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Link: https://lore.kernel.org/r/89c4a2487609a0ed6af3ecf01cc972bdc59a7a2d.1591634956.git.mirq-linux@rere.qmqm.pl
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: hda/realtek - add a pintbl quirk for several Lenovo machines
Hui Wang [Mon, 8 Jun 2020 11:55:41 +0000 (19:55 +0800)]
ALSA: hda/realtek - add a pintbl quirk for several Lenovo machines

commit 573fcbfd319ccef26caa3700320242accea7fd5c upstream.

A couple of Lenovo ThinkCentre machines all have 2 front mics and they
use the same codec alc623 and have the same pin config, so add a
pintbl entry for those machines to apply the fixup
ALC283_FIXUP_HEADSET_MIC.

Cc: <stable@vger.kernel.org>
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Link: https://lore.kernel.org/r/20200608115541.9531-1-hui.wang@canonical.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: fireface: fix configuration error for nominal sampling transfer frequency
Takashi Sakamoto [Sun, 10 May 2020 07:42:56 +0000 (16:42 +0900)]
ALSA: fireface: fix configuration error for nominal sampling transfer frequency

commit bbd6aac3ae15bef762af03bf62e35ace5c4292bd upstream.

128000 and 192000 are congruence modulo 32000, thus it's wrong to
distinguish them as multiple of 32000 and 48000 by modulo 32000 at
first.

Additionally, used condition statement to detect quadruple speed can
cause missing bit flag.

Furthermore, counter to ensure the configuration is wrong and it
causes false positive.

This commit fixes the above three bugs.

Cc: <stable@vger.kernel.org>
Fixes: 60aec494b389 ("ALSA: fireface: support allocate_resources operation in latter protocol")
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20200510074301.116224-2-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: es1688: Add the missed snd_card_free()
Chuhong Yuan [Wed, 3 Jun 2020 09:24:59 +0000 (17:24 +0800)]
ALSA: es1688: Add the missed snd_card_free()

commit d9b8fbf15d05350b36081eddafcf7b15aa1add50 upstream.

snd_es968_pnp_detect() misses a snd_card_free() in a failed path.
Add the missed function call to fix it.

Fixes: a20971b201ac ("ALSA: Merge es1688 and es968 drivers")
Signed-off-by: Chuhong Yuan <hslester96@gmail.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20200603092459.1424093-1-hslester96@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agowatchdog: imx_sc_wdt: Fix reboot on crash
Fabio Estevam [Sun, 12 Apr 2020 23:01:22 +0000 (20:01 -0300)]
watchdog: imx_sc_wdt: Fix reboot on crash

commit e56d48e92b1017b6a8dbe64923a889283733fd96 upstream.

Currently when running the samples/watchdog/watchdog-simple.c
application and forcing a kernel crash by doing:

# ./watchdog-simple &
# echo c > /proc/sysrq-trigger

The system does not reboot as expected.

Fix it by calling imx_sc_wdt_set_timeout() to configure the i.MX8QXP
watchdog with a proper timeout.

Cc: <stable@vger.kernel.org>
Fixes: 986857acbc9a ("watchdog: imx_sc: Add i.MX system controller watchdog support")
Reported-by: Breno Lima <breno.lima@nxp.com>
Signed-off-by: Fabio Estevam <festevam@gmail.com>
Reviewed-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Breno Lima <breno.lima@nxp.com>
Link: https://lore.kernel.org/r/20200412230122.5601-1-festevam@gmail.com
Signed-off-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Wim Van Sebroeck <wim@linux-watchdog.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agosmb3: add indatalen that can be a non-zero value to calculation of credit charge...
Namjae Jeon [Thu, 11 Jun 2020 02:21:19 +0000 (11:21 +0900)]
smb3: add indatalen that can be a non-zero value to calculation of credit charge in smb2 ioctl

commit ebf57440ec59a36e1fc5fe91e31d66ae0d1662d0 upstream.

Some of tests in xfstests failed with cifsd kernel server since commit
e80ddeb2f70e. cifsd kernel server validates credit charge from client
by calculating it base on max((InputCount + OutputCount) and
(MaxInputResponse + MaxOutputResponse)) according to specification.

MS-SMB2 specification describe credit charge calculation of smb2 ioctl :

If Connection.SupportsMultiCredit is TRUE, the server MUST validate
CreditCharge based on the maximum of (InputCount + OutputCount) and
(MaxInputResponse + MaxOutputResponse), as specified in section 3.3.5.2.5.
If the validation fails, it MUST fail the IOCTL request with
STATUS_INVALID_PARAMETER.

This patch add indatalen that can be a non-zero value to calculation of
credit charge in SMB2_ioctl_init().

Fixes: e80ddeb2f70e ("smb3: fix incorrect number of credits when ioctl
MaxOutputResponse > 64K")
Cc: Stable <stable@vger.kernel.org>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Cc: Steve French <smfrench@gmail.com>
Signed-off-by: Namjae Jeon <namjae.jeon@samsung.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agosmb3: fix incorrect number of credits when ioctl MaxOutputResponse > 64K
Steve French [Wed, 3 Jun 2020 06:33:58 +0000 (01:33 -0500)]
smb3: fix incorrect number of credits when ioctl MaxOutputResponse > 64K

commit e80ddeb2f70ebd0786aa7cdba3e58bc931fa0bb5 upstream.

We were not checking to see if ioctl requests asked for more than
64K (ie when CIFSMaxBufSize was > 64K) so when setting larger
CIFSMaxBufSize then ioctls would fail with invalid parameter errors.
When requests ask for more than 64K in MaxOutputResponse then we
need to ask for more than 1 credit.

Signed-off-by: Steve French <stfrench@microsoft.com>
CC: Stable <stable@vger.kernel.org>
Reviewed-by: Aurelien Aptel <aaptel@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoefi/efivars: Add missing kobject_put() in sysfs entry creation error path
Ard Biesheuvel [Fri, 22 May 2020 16:15:49 +0000 (18:15 +0200)]
efi/efivars: Add missing kobject_put() in sysfs entry creation error path

commit d8bd8c6e2cfab8b78b537715255be8d7557791c0 upstream.

The documentation provided by kobject_init_and_add() clearly spells out
the need to call kobject_put() on the kobject if an error is returned.
Add this missing call to the error path.

Cc: <stable@vger.kernel.org>
Reported-by: 亿一 <teroincn@gmail.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoio_uring: use kvfree() in io_sqe_buffer_register()
Denis Efremov [Fri, 5 Jun 2020 09:32:03 +0000 (12:32 +0300)]
io_uring: use kvfree() in io_sqe_buffer_register()

commit a8c73c1a614f6da6c0b04c393f87447e28cb6de4 upstream.

Use kvfree() to free the pages and vmas, since they are allocated by
kvmalloc_array() in a loop.

Fixes: d4ef647510b1 ("io_uring: avoid page allocation warnings")
Signed-off-by: Denis Efremov <efremov@linux.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200605093203.40087-1-efremov@linux.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoASoC: max9867: fix volume controls
Pavel Dobias [Fri, 15 May 2020 12:07:57 +0000 (14:07 +0200)]
ASoC: max9867: fix volume controls

commit 8ba4dc3cff8cbe2c571063a5fd7116e8bde563ca upstream.

The xmax values for Master Playback Volume and Mic Boost
Capture Volume are specified incorrectly (one greater)
which results in the wrong dB gain being shown to the user
in the case of Master Playback Volume.

Signed-off-by: Pavel Dobias <dobias@2n.cz>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20200515120757.24669-1-dobias@2n.cz
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agopowerpc/ptdump: Properly handle non standard page size
Christophe Leroy [Tue, 19 May 2020 05:48:54 +0000 (05:48 +0000)]
powerpc/ptdump: Properly handle non standard page size

commit b00ff6d8c1c3898b0f768cbb38ef722d25bd2f39 upstream.

In order to properly display information regardless of the page size,
it is necessary to take into account real page size.

Fixes: cabe8138b23c ("powerpc: dump as a single line areas mapping a single physical page.")
Cc: stable@vger.kernel.org
Signed-off-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/a53b2a0ffd042a8d85464bf90d55bc5b970e00a1.1589866984.git.christophe.leroy@csgroup.eu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: x86: Fix APIC page invalidation race
Eiichi Tsukata [Sat, 6 Jun 2020 04:26:27 +0000 (13:26 +0900)]
KVM: x86: Fix APIC page invalidation race

commit e649b3f0188f8fd34dd0dde8d43fd3312b902fb2 upstream.

Commit b1394e745b94 ("KVM: x86: fix APIC page invalidation") tried
to fix inappropriate APIC page invalidation by re-introducing arch
specific kvm_arch_mmu_notifier_invalidate_range() and calling it from
kvm_mmu_notifier_invalidate_range_start. However, the patch left a
possible race where the VMCS APIC address cache is updated *before*
it is unmapped:

  (Invalidator) kvm_mmu_notifier_invalidate_range_start()
  (Invalidator) kvm_make_all_cpus_request(kvm, KVM_REQ_APIC_PAGE_RELOAD)
  (KVM VCPU) vcpu_enter_guest()
  (KVM VCPU) kvm_vcpu_reload_apic_access_page()
  (Invalidator) actually unmap page

Because of the above race, there can be a mismatch between the
host physical address stored in the APIC_ACCESS_PAGE VMCS field and
the host physical address stored in the EPT entry for the APIC GPA
(0xfee0000).  When this happens, the processor will not trap APIC
accesses, and will instead show the raw contents of the APIC-access page.
Because Windows OS periodically checks for unexpected modifications to
the LAPIC register, this will show up as a BSOD crash with BugCheck
CRITICAL_STRUCTURE_CORRUPTION (109) we are currently seeing in
https://bugzilla.redhat.com/show_bug.cgi?id=1751017.

The root cause of the issue is that kvm_arch_mmu_notifier_invalidate_range()
cannot guarantee that no additional references are taken to the pages in
the range before kvm_mmu_notifier_invalidate_range_end().  Fortunately,
this case is supported by the MMU notifier API, as documented in
include/linux/mmu_notifier.h:

 * If the subsystem
         * can't guarantee that no additional references are taken to
         * the pages in the range, it has to implement the
         * invalidate_range() notifier to remove any references taken
         * after invalidate_range_start().

The fix therefore is to reload the APIC-access page field in the VMCS
from kvm_mmu_notifier_invalidate_range() instead of ..._range_start().

Cc: stable@vger.kernel.org
Fixes: b1394e745b94 ("KVM: x86: fix APIC page invalidation")
Fixes: https://bugzilla.kernel.org/show_bug.cgi?id=197951
Signed-off-by: Eiichi Tsukata <eiichi.tsukata@nutanix.com>
Message-Id: <20200606042627.61070-1-eiichi.tsukata@nutanix.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: x86: respect singlestep when emulating instruction
Felipe Franciosi [Tue, 19 May 2020 08:11:22 +0000 (08:11 +0000)]
KVM: x86: respect singlestep when emulating instruction

commit 384dea1c9183880be183cfaae161d99aafd16df6 upstream.

When userspace configures KVM_GUESTDBG_SINGLESTEP, KVM will manage the
presence of X86_EFLAGS_TF via kvm_set/get_rflags on vcpus. The actual
rflag bit is therefore hidden from callers.

That includes init_emulate_ctxt() which uses the value returned from
kvm_get_flags() to set ctxt->tf. As a result, x86_emulate_instruction()
will skip a single step, leaving singlestep_rip stale and not returning
to userspace.

This resolves the issue by observing the vcpu guest_debug configuration
alongside ctxt->tf in x86_emulate_instruction(), performing the single
step if set.

Cc: stable@vger.kernel.org
Signed-off-by: Felipe Franciosi <felipe@nutanix.com>
Message-Id: <20200519081048.8204-1-felipe@nutanix.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: x86/mmu: Set mmio_value to '0' if reserved #PF can't be generated
Sean Christopherson [Wed, 27 May 2020 08:49:09 +0000 (01:49 -0700)]
KVM: x86/mmu: Set mmio_value to '0' if reserved #PF can't be generated

commit 6129ed877d409037b79866327102c9dc59a302fe upstream.

Set the mmio_value to '0' instead of simply clearing the present bit to
squash a benign warning in kvm_mmu_set_mmio_spte_mask() that complains
about the mmio_value overlapping the lower GFN mask on systems with 52
bits of PA space.

Opportunistically clean up the code and comments.

Cc: stable@vger.kernel.org
Fixes: d43e2675e96fc ("KVM: x86: only do L1TF workaround on affected processors")
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20200527084909.23492-1-sean.j.christopherson@intel.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoperf/x86/intel: Add more available bits for OFFCORE_RESPONSE of Intel Tremont
Kan Liang [Fri, 1 May 2020 12:54:42 +0000 (05:54 -0700)]
perf/x86/intel: Add more available bits for OFFCORE_RESPONSE of Intel Tremont

commit 0813c40556fce1eeefb996e020cc5339e0b84137 upstream.

The mask in the extra_regs for Intel Tremont need to be extended to
allow more defined bits.

"Outstanding Requests" (bit 63) is only available on MSR_OFFCORE_RSP0;

Fixes: 6daeb8737f8a ("perf/x86/intel: Add Tremont core PMU support")
Reported-by: Stephane Eranian <eranian@google.com>
Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20200501125442.7030-1-kan.liang@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agox86/reboot/quirks: Add MacBook6,1 reboot quirk
Hill Ma [Sat, 25 Apr 2020 20:06:41 +0000 (13:06 -0700)]
x86/reboot/quirks: Add MacBook6,1 reboot quirk

commit 140fd4ac78d385e6c8e6a5757585f6c707085f87 upstream.

On MacBook6,1 reboot would hang unless parameter reboot=pci is added.
Make it automatic.

Signed-off-by: Hill Ma <maahiuzeon@gmail.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: stable@vger.kernel.org
Link: https://lkml.kernel.org/r/20200425200641.GA1554@cslab.localdomain
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agox86/speculation: PR_SPEC_FORCE_DISABLE enforcement for indirect branches.
Anthony Steinhauser [Sun, 7 Jun 2020 12:44:19 +0000 (05:44 -0700)]
x86/speculation: PR_SPEC_FORCE_DISABLE enforcement for indirect branches.

commit 4d8df8cbb9156b0a0ab3f802b80cb5db57acc0bf upstream.

Currently, it is possible to enable indirect branch speculation even after
it was force-disabled using the PR_SPEC_FORCE_DISABLE option. Moreover, the
PR_GET_SPECULATION_CTRL command gives afterwards an incorrect result
(force-disabled when it is in fact enabled). This also is inconsistent
vs. STIBP and the documention which cleary states that
PR_SPEC_FORCE_DISABLE cannot be undone.

Fix this by actually enforcing force-disabled indirect branch
speculation. PR_SPEC_ENABLE called after PR_SPEC_FORCE_DISABLE now fails
with -EPERM as described in the documentation.

Fixes: 9137bb27e60e ("x86/speculation: Add prctl() control for indirect branch speculation")
Signed-off-by: Anthony Steinhauser <asteinhauser@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agox86/speculation: Avoid force-disabling IBPB based on STIBP and enhanced IBRS.
Anthony Steinhauser [Tue, 19 May 2020 13:40:42 +0000 (06:40 -0700)]
x86/speculation: Avoid force-disabling IBPB based on STIBP and enhanced IBRS.

commit 21998a351512eba4ed5969006f0c55882d995ada upstream.

When STIBP is unavailable or enhanced IBRS is available, Linux
force-disables the IBPB mitigation of Spectre-BTB even when simultaneous
multithreading is disabled. While attempts to enable IBPB using
prctl(PR_SET_SPECULATION_CTRL, PR_SPEC_INDIRECT_BRANCH, ...) fail with
EPERM, the seccomp syscall (or its prctl(PR_SET_SECCOMP, ...) equivalent)
which are used e.g. by Chromium or OpenSSH succeed with no errors but the
application remains silently vulnerable to cross-process Spectre v2 attacks
(classical BTB poisoning). At the same time the SYSFS reporting
(/sys/devices/system/cpu/vulnerabilities/spectre_v2) displays that IBPB is
conditionally enabled when in fact it is unconditionally disabled.

STIBP is useful only when SMT is enabled. When SMT is disabled and STIBP is
unavailable, it makes no sense to force-disable also IBPB, because IBPB
protects against cross-process Spectre-BTB attacks regardless of the SMT
state. At the same time since missing STIBP was only observed on AMD CPUs,
AMD does not recommend using STIBP, but recommends using IBPB, so disabling
IBPB because of missing STIBP goes directly against AMD's advice:
https://developer.amd.com/wp-content/resources/Architecture_Guidelines_Update_Indirect_Branch_Control.pdf

Similarly, enhanced IBRS is designed to protect cross-core BTB poisoning
and BTB-poisoning attacks from user space against kernel (and
BTB-poisoning attacks from guest against hypervisor), it is not designed
to prevent cross-process (or cross-VM) BTB poisoning between processes (or
VMs) running on the same core. Therefore, even with enhanced IBRS it is
necessary to flush the BTB during context-switches, so there is no reason
to force disable IBPB when enhanced IBRS is available.

Enable the prctl control of IBPB even when STIBP is unavailable or enhanced
IBRS is available.

Fixes: 7cc765a67d8e ("x86/speculation: Enable prctl mode for spectre_v2_user")
Signed-off-by: Anthony Steinhauser <asteinhauser@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agox86/speculation: Prevent rogue cross-process SSBD shutdown
Anthony Steinhauser [Sun, 5 Jan 2020 20:19:43 +0000 (12:19 -0800)]
x86/speculation: Prevent rogue cross-process SSBD shutdown

commit dbbe2ad02e9df26e372f38cc3e70dab9222c832e upstream.

On context switch the change of TIF_SSBD and TIF_SPEC_IB are evaluated
to adjust the mitigations accordingly. This is optimized to avoid the
expensive MSR write if not needed.

This optimization is buggy and allows an attacker to shutdown the SSBD
protection of a victim process.

The update logic reads the cached base value for the speculation control
MSR which has neither the SSBD nor the STIBP bit set. It then OR's the
SSBD bit only when TIF_SSBD is different and requests the MSR update.

That means if TIF_SSBD of the previous and next task are the same, then
the base value is not updated, even if TIF_SSBD is set. The MSR write is
not requested.

Subsequently if the TIF_STIBP bit differs then the STIBP bit is updated
in the base value and the MSR is written with a wrong SSBD value.

This was introduced when the per task/process conditional STIPB
switching was added on top of the existing SSBD switching.

It is exploitable if the attacker creates a process which enforces SSBD
and has the contrary value of STIBP than the victim process (i.e. if the
victim process enforces STIBP, the attacker process must not enforce it;
if the victim process does not enforce STIBP, the attacker process must
enforce it) and schedule it on the same core as the victim process. If
the victim runs after the attacker the victim becomes vulnerable to
Spectre V4.

To fix this, update the MSR value independent of the TIF_SSBD difference
and dependent on the SSBD mitigation method available. This ensures that
a subsequent STIPB initiated MSR write has the correct state of SSBD.

[ tglx: Handle X86_FEATURE_VIRT_SSBD & X86_FEATURE_VIRT_SSBD correctly
        and massaged changelog ]

Fixes: 5bfbe3ad5840 ("x86/speculation: Prepare for per task indirect branch speculation control")
Signed-off-by: Anthony Steinhauser <asteinhauser@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agox86/PCI: Mark Intel C620 MROMs as having non-compliant BARs
Xiaochun Lee [Fri, 15 May 2020 03:31:07 +0000 (23:31 -0400)]
x86/PCI: Mark Intel C620 MROMs as having non-compliant BARs

commit 1574051e52cb4b5b7f7509cfd729b76ca1117808 upstream.

The Intel C620 Platform Controller Hub has MROM functions that have non-PCI
registers (undocumented in the public spec) where BAR 0 is supposed to be,
which results in messages like this:

  pci 0000:00:11.0: [Firmware Bug]: reg 0x30: invalid BAR (can't size)

Mark these MROM functions as having non-compliant BARs so we don't try to
probe any of them.  There are no other BARs on these devices.

See the Intel C620 Series Chipset Platform Controller Hub Datasheet,
May 2019, Document Number 336067-007US, sec 2.1, 35.5, 35.6.

[bhelgaas: commit log, add 0xa26d]
Link: https://lore.kernel.org/r/1589513467-17070-1-git-send-email-lixiaochun.2888@163.com
Signed-off-by: Xiaochun Lee <lixc17@lenovo.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agox86_64: Fix jiffies ODR violation
Bob Haarman [Tue, 2 Jun 2020 19:30:59 +0000 (12:30 -0700)]
x86_64: Fix jiffies ODR violation

commit d8ad6d39c35d2b44b3d48b787df7f3359381dcbf upstream.

'jiffies' and 'jiffies_64' are meant to alias (two different symbols that
share the same address).  Most architectures make the symbols alias to the
same address via a linker script assignment in their
arch/<arch>/kernel/vmlinux.lds.S:

jiffies = jiffies_64;

which is effectively a definition of jiffies.

jiffies and jiffies_64 are both forward declared for all architectures in
include/linux/jiffies.h. jiffies_64 is defined in kernel/time/timer.c.

x86_64 was peculiar in that it wasn't doing the above linker script
assignment, but rather was:
1. defining jiffies in arch/x86/kernel/time.c instead via the linker script.
2. overriding the symbol jiffies_64 from kernel/time/timer.c in
arch/x86/kernel/vmlinux.lds.s via 'jiffies_64 = jiffies;'.

As Fangrui notes:

  In LLD, symbol assignments in linker scripts override definitions in
  object files. GNU ld appears to have the same behavior. It would
  probably make sense for LLD to error "duplicate symbol" but GNU ld
  is unlikely to adopt for compatibility reasons.

This results in an ODR violation (UB), which seems to have survived
thus far. Where it becomes harmful is when;

1. -fno-semantic-interposition is used:

As Fangrui notes:

  Clang after LLVM commit 5b22bcc2b70d
  ("[X86][ELF] Prefer to lower MC_GlobalAddress operands to .Lfoo$local")
  defaults to -fno-semantic-interposition similar semantics which help
  -fpic/-fPIC code avoid GOT/PLT when the referenced symbol is defined
  within the same translation unit. Unlike GCC
  -fno-semantic-interposition, Clang emits such relocations referencing
  local symbols for non-pic code as well.

This causes references to jiffies to refer to '.Ljiffies$local' when
jiffies is defined in the same translation unit. Likewise, references to
jiffies_64 become references to '.Ljiffies_64$local' in translation units
that define jiffies_64.  Because these differ from the names used in the
linker script, they will not be rewritten to alias one another.

2. Full LTO

Full LTO effectively treats all source files as one translation
unit, causing these local references to be produced everywhere.  When
the linker processes the linker script, there are no longer any
references to jiffies_64' anywhere to replace with 'jiffies'.  And
thus '.Ljiffies$local' and '.Ljiffies_64$local' no longer alias
at all.

In the process of porting patches enabling Full LTO from arm64 to x86_64,
spooky bugs have been observed where the kernel appeared to boot, but init
doesn't get scheduled.

Avoid the ODR violation by matching other architectures and define jiffies
only by linker script.  For -fno-semantic-interposition + Full LTO, there
is no longer a global definition of jiffies for the compiler to produce a
local symbol which the linker script won't ensure aliases to jiffies_64.

Fixes: 40747ffa5aa8 ("asmlinkage: Make jiffies visible")
Reported-by: Nathan Chancellor <natechancellor@gmail.com>
Reported-by: Alistair Delva <adelva@google.com>
Debugged-by: Nick Desaulniers <ndesaulniers@google.com>
Debugged-by: Sami Tolvanen <samitolvanen@google.com>
Suggested-by: Fangrui Song <maskray@google.com>
Signed-off-by: Bob Haarman <inglorion@google.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Sedat Dilek <sedat.dilek@gmail.com> # build+boot on
Reviewed-by: Andi Kleen <ak@linux.intel.com>
Reviewed-by: Josh Poimboeuf <jpoimboe@redhat.com>
Cc: stable@vger.kernel.org
Link: https://github.com/ClangBuiltLinux/linux/issues/852
Link: https://lkml.kernel.org/r/20200602193100.229287-1-inglorion@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agosctp: fix refcount bug in sctp_wfree
Qiujun Huang [Fri, 27 Mar 2020 03:07:51 +0000 (11:07 +0800)]
sctp: fix refcount bug in sctp_wfree

[ Upstream commit 5c3e82fe159622e46e91458c1a6509c321a62820 ]

We should iterate over the datamsgs to move
all chunks(skbs) to newsk.

The following case cause the bug:
for the trouble SKB, it was in outq->transmitted list

sctp_outq_sack
        sctp_check_transmitted
                SKB was moved to outq->sacked list
        then throw away the sack queue
                SKB was deleted from outq->sacked
(but it was held by datamsg at sctp_datamsg_to_asoc
So, sctp_wfree was not called here)

then migrate happened

        sctp_for_each_tx_datachunk(
        sctp_clear_owner_w);
        sctp_assoc_migrate();
        sctp_for_each_tx_datachunk(
        sctp_set_owner_w);
SKB was not in the outq, and was not changed to newsk

finally

__sctp_outq_teardown
        sctp_chunk_put (for another skb)
                sctp_datamsg_put
                        __kfree_skb(msg->frag_list)
                                sctp_wfree (for SKB)
SKB->sk was still oldsk (skb->sk != asoc->base.sk).

Reported-and-tested-by: syzbot+cea71eec5d6de256d54d@syzkaller.appspotmail.com
Signed-off-by: Qiujun Huang <hqjagain@gmail.com>
Acked-by: Marcelo Ricardo Leitner <mleitner@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agosctp: fix possibly using a bad saddr with a given dst
Marcelo Ricardo Leitner [Thu, 26 Mar 2020 23:47:46 +0000 (20:47 -0300)]
sctp: fix possibly using a bad saddr with a given dst

[ Upstream commit 582eea230536a6f104097dd46205822005d5fe3a ]

Under certain circumstances, depending on the order of addresses on the
interfaces, it could be that sctp_v[46]_get_dst() would return a dst
with a mismatched struct flowi.

For example, if when walking through the bind addresses and the first
one is not a match, it saves the dst as a fallback (added in
410f03831c07), but not the flowi. Then if the next one is also not a
match, the previous dst will be returned but with the flowi information
for the 2nd address, which is wrong.

The fix is to use a locally stored flowi that can be used for such
attempts, and copy it to the parameter only in case it is a possible
match, together with the corresponding dst entry.

The patch updates IPv6 code mostly just to be in sync. Even though the issue
is also present there, it fallback is not expected to work with IPv6.

Fixes: 410f03831c07 ("sctp: add routing output fallback")
Reported-by: Jin Meng <meng.a.jin@nokia-sbell.com>
Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Tested-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoiwlwifi: mvm: fix NVM check for 3168 devices
Luca Coelho [Mon, 25 Nov 2019 11:21:58 +0000 (13:21 +0200)]
iwlwifi: mvm: fix NVM check for 3168 devices

[ Upstream commit b3f20e098293892388d6a0491d6bbb2efb46fbff ]

We had a check on !NVM_EXT and then a check for NVM_SDP in the else
block of this if.  The else block, obviously, could only be reached if
using NVM_EXT, so it would never be NVM_SDP.

Fix that by checking whether the nvm_type is IWL_NVM instead of
checking for !IWL_NVM_EXT to solve this issue.

Reported-by: Stefan Sperling <stsp@stsp.name>
Signed-off-by: Luca Coelho <luciano.coelho@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoipv4: fix a RCU-list lock in fib_triestat_seq_show
Qian Cai [Wed, 25 Mar 2020 22:01:00 +0000 (18:01 -0400)]
ipv4: fix a RCU-list lock in fib_triestat_seq_show

[ Upstream commit fbe4e0c1b298b4665ee6915266c9d6c5b934ef4a ]

fib_triestat_seq_show() calls hlist_for_each_entry_rcu(tb, head,
tb_hlist) without rcu_read_lock() will trigger a warning,

 net/ipv4/fib_trie.c:2579 RCU-list traversed in non-reader section!!

 other info that might help us debug this:

 rcu_scheduler_active = 2, debug_locks = 1
 1 lock held by proc01/115277:
  #0: c0000014507acf00 (&p->lock){+.+.}-{3:3}, at: seq_read+0x58/0x670

 Call Trace:
  dump_stack+0xf4/0x164 (unreliable)
  lockdep_rcu_suspicious+0x140/0x164
  fib_triestat_seq_show+0x750/0x880
  seq_read+0x1a0/0x670
  proc_reg_read+0x10c/0x1b0
  __vfs_read+0x3c/0x70
  vfs_read+0xac/0x170
  ksys_read+0x7c/0x140
  system_call+0x5c/0x68

Fix it by adding a pair of rcu_read_lock/unlock() and use
cond_resched_rcu() to avoid the situation where walking of a large
number of items  may prevent scheduling for a long time.

Signed-off-by: Qian Cai <cai@lca.pw>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoaio: fix async fsync creds
Miklos Szeredi [Thu, 14 May 2020 14:44:24 +0000 (16:44 +0200)]
aio: fix async fsync creds

commit 530f32fc370fd1431ea9802dbc53ab5601dfccdb upstream.

Avi Kivity reports that on fuse filesystems running in a user namespace
asyncronous fsync fails with EOVERFLOW.

The reason is that f_ops->fsync() is called with the creds of the kthread
performing aio work instead of the creds of the process originally
submitting IOCB_CMD_FSYNC.

Fuse sends the creds of the caller in the request header and it needs to
translate the uid and gid into the server's user namespace.  Since the
kthread is running in init_user_ns, the translation will fail and the
operation returns an error.

It can be argued that fsync doesn't actually need any creds, but just
zeroing out those fields in the header (as with requests that currently
don't take creds) is a backward compatibility risk.

Instead of working around this issue in fuse, solve the core of the problem
by calling the filesystem with the proper creds.

Reported-by: Avi Kivity <avi@scylladb.com>
Tested-by: Giuseppe Scrivano <gscrivan@redhat.com>
Fixes: c9582eb0ff7d ("fuse: Fail all requests with invalid uids or gids")
Cc: stable@vger.kernel.org # 4.18+
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agofanotify: fix ignore mask logic for events on child and on dir
Amir Goldstein [Sun, 24 May 2020 07:24:41 +0000 (10:24 +0300)]
fanotify: fix ignore mask logic for events on child and on dir

[ Upstream commit 2f02fd3fa13e51713b630164f8a8e5b42de8283b ]

The comments in fanotify_group_event_mask() say:

  "If the event is on dir/child and this mark doesn't care about
   events on dir/child, don't send it!"

Specifically, mount and filesystem marks do not care about events
on child, but they can still specify an ignore mask for those events.
For example, a group that has:
- A mount mark with mask 0 and ignore_mask FAN_OPEN
- An inode mark on a directory with mask FAN_OPEN | FAN_OPEN_EXEC
  with flag FAN_EVENT_ON_CHILD

A child file open for exec would be reported to group with the FAN_OPEN
event despite the fact that FAN_OPEN is in ignore mask of mount mark,
because the mark iteration loop skips over non-inode marks for events
on child when calculating the ignore mask.

Move ignore mask calculation to the top of the iteration loop block
before excluding marks for events on dir/child.

Link: https://lore.kernel.org/r/20200524072441.18258-1-amir73il@gmail.com
Reported-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/linux-fsdevel/20200521162443.GA26052@quack2.suse.cz/
Fixes: 55bf882c7f13 "fanotify: fix merging marks masks with FAN_ONDIR"
Fixes: b469e7e47c8a "fanotify: fix handling of events on child..."
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Jan Kara <jack@suse.cz>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoselftests: fix flower parent qdisc
Vlad Buslov [Thu, 14 May 2020 06:35:52 +0000 (09:35 +0300)]
selftests: fix flower parent qdisc

[ Upstream commit 0531b0357ba37464e5c0033e1b7c69bbf5ecd8fb ]

Flower tests used to create ingress filter with specified parent qdisc
"parent ffff:" but dump them on "ingress". With recent commit that fixed
tcm_parent handling in dump those are not considered same parent anymore,
which causes iproute2 tc to emit additional "parent ffff:" in first line of
filter dump output. The change in output causes filter match in tests to
fail.

Prevent parent qdisc output when dumping filters in flower tests by always
correctly specifying "ingress" parent both when creating and dumping
filters.

Fixes: a7df4870d79b ("net_sched: fix tcm_parent in tc filter dump")
Signed-off-by: Vlad Buslov <vladbu@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agomm: add kvfree_sensitive() for freeing sensitive data objects
Waiman Long [Thu, 4 Jun 2020 23:48:21 +0000 (16:48 -0700)]
mm: add kvfree_sensitive() for freeing sensitive data objects

[ Upstream commit d4eaa2837851db2bfed572898bfc17f9a9f9151e ]

For kvmalloc'ed data object that contains sensitive information like
cryptographic keys, we need to make sure that the buffer is always cleared
before freeing it.  Using memset() alone for buffer clearing may not
provide certainty as the compiler may compile it away.  To be sure, the
special memzero_explicit() has to be used.

This patch introduces a new kvfree_sensitive() for freeing those sensitive
data objects allocated by kvmalloc().  The relevant places where
kvfree_sensitive() can be used are modified to use it.

Fixes: 4f0882491a14 ("KEYS: Avoid false positive ENOMEM error on key read")
Suggested-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Acked-by: David Howells <dhowells@redhat.com>
Cc: Jarkko Sakkinen <jarkko.sakkinen@linux.intel.com>
Cc: James Morris <jmorris@namei.org>
Cc: "Serge E. Hallyn" <serge@hallyn.com>
Cc: Joe Perches <joe@perches.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Uladzislau Rezki <urezki@gmail.com>
Link: http://lkml.kernel.org/r/20200407200318.11711-1-longman@redhat.com
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoperf probe: Accept the instance number of kretprobe event
Masami Hiramatsu [Wed, 6 May 2020 14:29:12 +0000 (23:29 +0900)]
perf probe: Accept the instance number of kretprobe event

[ Upstream commit c6aab66a728b6518772c74bd9dff66e1a1c652fd ]

Since the commit 6a13a0d7b4d1 ("ftrace/kprobe: Show the maxactive number
on kprobe_events") introduced to show the instance number of kretprobe
events, the length of the 1st format of the kprobe event will not 1, but
it can be longer.  This caused a parser error in perf-probe.

Skip the length check the 1st format of the kprobe event to accept this
instance number.

Without this fix:

  # perf probe -a vfs_read%return
  Added new event:
    probe:vfs_read__return (on vfs_read%return)

  You can now use it in all perf tools, such as:

   perf record -e probe:vfs_read__return -aR sleep 1

  # perf probe -l
  Semantic error :Failed to parse event name: r16:probe/vfs_read__return
    Error: Failed to show event list.

And with this fixes:

  # perf probe -a vfs_read%return
  ...
  # perf probe -l
    probe:vfs_read__return (on vfs_read%return)

Fixes: 6a13a0d7b4d1 ("ftrace/kprobe: Show the maxactive number on kprobe_events")
Reported-by: Yuxuan Shui <yshuiv7@gmail.com>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Tested-by: Yuxuan Shui <yshuiv7@gmail.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: stable@vger.kernel.org
Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=207587
Link: http://lore.kernel.org/lkml/158877535215.26469.1113127926699134067.stgit@devnote2
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoPCI/PM: Adjust pcie_wait_for_link_delay() for caller delay
Bjorn Helgaas [Fri, 15 May 2020 19:31:16 +0000 (14:31 -0500)]
PCI/PM: Adjust pcie_wait_for_link_delay() for caller delay

[ Upstream commit f044baaff1eb7ae5aa7a36f1b7ad5bd8eeb672c4 ]

The caller of pcie_wait_for_link_delay() specifies the time to wait after
the link becomes active.  When the downstream port doesn't support link
active reporting, obviously we can't tell when the link becomes active, so
we waited the worst-case time (1000 ms) plus 100 ms, ignoring the delay
from the caller.

Instead, wait for 1000 ms + the delay from the caller.

Fixes: 4827d63891b6 ("PCI/PM: Add pcie_wait_for_link_delay()")
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoKVM: x86: only do L1TF workaround on affected processors
Paolo Bonzini [Tue, 19 May 2020 09:34:41 +0000 (05:34 -0400)]
KVM: x86: only do L1TF workaround on affected processors

[ Upstream commit d43e2675e96fc6ae1a633b6a69d296394448cc32 ]

KVM stores the gfn in MMIO SPTEs as a caching optimization.  These are split
in two parts, as in "[high 11111 low]", to thwart any attempt to use these bits
in an L1TF attack.  This works as long as there are 5 free bits between
MAXPHYADDR and bit 50 (inclusive), leaving bit 51 free so that the MMIO
access triggers a reserved-bit-set page fault.

The bit positions however were computed wrongly for AMD processors that have
encryption support.  In this case, x86_phys_bits is reduced (for example
from 48 to 43, to account for the C bit at position 47 and four bits used
internally to store the SEV ASID and other stuff) while x86_cache_bits in
would remain set to 48, and _all_ bits between the reduced MAXPHYADDR
and bit 51 are set.  Then low_phys_bits would also cover some of the
bits that are set in the shadow_mmio_value, terribly confusing the gfn
caching mechanism.

To fix this, avoid splitting gfns as long as the processor does not have
the L1TF bug (which includes all AMD processors).  When there is no
splitting, low_phys_bits can be set to the reduced MAXPHYADDR removing
the overlap.  This fixes "npt=0" operation on EPYC processors.

Thanks to Maxim Levitsky for bisecting this bug.

Cc: stable@vger.kernel.org
Fixes: 52918ed5fcf0 ("KVM: SVM: Override default MMIO mask if memory encryption is enabled")
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agox86/cpu/amd: Make erratum #1054 a legacy erratum
Kim Phillips [Fri, 17 Apr 2020 14:33:56 +0000 (09:33 -0500)]
x86/cpu/amd: Make erratum #1054 a legacy erratum

[ Upstream commit e2abfc0448a46d8a137505aa180caf14070ec535 ]

Commit

  21b5ee59ef18 ("x86/cpu/amd: Enable the fixed Instructions Retired
 counter IRPERF")

mistakenly added erratum #1054 as an OS Visible Workaround (OSVW) ID 0.
Erratum #1054 is not OSVW ID 0 [1], so make it a legacy erratum.

There would never have been a false positive on older hardware that
has OSVW bit 0 set, since the IRPERF feature was not available.

However, save a couple of RDMSR executions per thread, on modern
system configurations that correctly set non-zero values in their
OSVW_ID_Length MSRs.

[1] Revision Guide for AMD Family 17h Models 00h-0Fh Processors. The
revision guide is available from the bugzilla link below.

Fixes: 21b5ee59ef18 ("x86/cpu/amd: Enable the fixed Instructions Retired counter IRPERF")
Reported-by: Andrew Cooper <andrew.cooper3@citrix.com>
Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lkml.kernel.org/r/20200417143356.26054-1-kim.phillips@amd.com
Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agos390/pci: Log new handle in clp_disable_fh()
Petr Tesarik [Fri, 22 May 2020 18:39:22 +0000 (20:39 +0200)]
s390/pci: Log new handle in clp_disable_fh()

[ Upstream commit e1750a3d9abbea2ece29cac8dc5a6f5bc19c1492 ]

After disabling a function, the original handle is logged instead of
the disabled handle.

Link: https://lkml.kernel.org/r/20200522183922.5253-1-ptesarik@suse.com
Fixes: 17cdec960cf7 ("s390/pci: Recover handle in clp_set_pci_fn()")
Reviewed-by: Pierre Morel <pmorel@linux.ibm.com>
Signed-off-by: Petr Tesarik <ptesarik@suse.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agopadata: add separate cpuhp node for CPUHP_PADATA_DEAD
Daniel Jordan [Tue, 21 Apr 2020 16:34:55 +0000 (12:34 -0400)]
padata: add separate cpuhp node for CPUHP_PADATA_DEAD

[ Upstream commit 3c2214b6027ff37945799de717c417212e1a8c54 ]

Removing the pcrypt module triggers this:

  general protection fault, probably for non-canonical
    address 0xdead000000000122
  CPU: 5 PID: 264 Comm: modprobe Not tainted 5.6.0+ #2
  Hardware name: QEMU Standard PC
  RIP: 0010:__cpuhp_state_remove_instance+0xcc/0x120
  Call Trace:
   padata_sysfs_release+0x74/0xce
   kobject_put+0x81/0xd0
   padata_free+0x12/0x20
   pcrypt_exit+0x43/0x8ee [pcrypt]

padata instances wrongly use the same hlist node for the online and dead
states, so __padata_free()'s second cpuhp remove call chokes on the node
that the first poisoned.

cpuhp multi-instance callbacks only walk forward in cpuhp_step->list and
the same node is linked in both the online and dead lists, so the list
corruption that results from padata_alloc() adding the node to a second
list without removing it from the first doesn't cause problems as long
as no instances are freed.

Avoid the issue by giving each state its own node.

Fixes: 894c9ef9780c ("padata: validate cpumask without removed CPU during offline")
Signed-off-by: Daniel Jordan <daniel.m.jordan@oracle.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Steffen Klassert <steffen.klassert@secunet.com>
Cc: linux-crypto@vger.kernel.org
Cc: linux-kernel@vger.kernel.org
Cc: stable@vger.kernel.org # v5.4+
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoRDMA/uverbs: Make the event_queue fds return POLLERR when disassociated
Jason Gunthorpe [Tue, 7 Apr 2020 00:44:26 +0000 (21:44 -0300)]
RDMA/uverbs: Make the event_queue fds return POLLERR when disassociated

[ Upstream commit eb356e6dc15a30af604f052cd0e170450193c254 ]

If is_closed is set, and the event list is empty, then read() will return
-EIO without blocking. After setting is_closed in
ib_uverbs_free_event_queue(), we do trigger a wake_up on the poll_wait,
but the fops->poll() function does not check it, so poll will continue to
sleep on an empty list.

Fixes: 14e23bd6d221 ("RDMA/core: Fix locking in ib_uverbs_event_read")
Link: https://lore.kernel.org/r/0-v1-ace813388969+48859-uverbs_poll_fix%25jgg@mellanox.com
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agosmack: avoid unused 'sip' variable warning
Arnd Bergmann [Wed, 8 Apr 2020 19:04:31 +0000 (21:04 +0200)]
smack: avoid unused 'sip' variable warning

[ Upstream commit 00720f0e7f288d29681d265c23b22bb0f0f4e5b4 ]

The mix of IS_ENABLED() and #ifdef checks has left a combination
that causes a warning about an unused variable:

security/smack/smack_lsm.c: In function 'smack_socket_connect':
security/smack/smack_lsm.c:2838:24: error: unused variable 'sip' [-Werror=unused-variable]
 2838 |   struct sockaddr_in6 *sip = (struct sockaddr_in6 *)sap;

Change the code to use C-style checks consistently so the compiler
can handle it correctly.

Fixes: 87fbfffcc89b ("broken ping to ipv6 linklocal addresses on debian buster")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Casey Schaufler <casey@schaufler-ca.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoath9k_htc: Silence undersized packet warnings
Masashi Honma [Mon, 4 May 2020 21:44:43 +0000 (06:44 +0900)]
ath9k_htc: Silence undersized packet warnings

[ Upstream commit 450edd2805982d14ed79733a82927d2857b27cac ]

Some devices like TP-Link TL-WN722N produces this kind of messages
frequently.

kernel: ath: phy0: Short RX data len, dropping (dlen: 4)

This warning is useful for developers to recognize that the device
(Wi-Fi dongle or USB hub etc) is noisy but not for general users. So
this patch make this warning to debug message.

Reported-By: Denis <pro.denis@protonmail.com>
Ref: https://bugzilla.kernel.org/show_bug.cgi?id=207539
Fixes: cd486e627e67 ("ath9k_htc: Discard undersized packets")
Signed-off-by: Masashi Honma <masashi.honma@gmail.com>
Signed-off-by: Kalle Valo <kvalo@codeaurora.org>
Link: https://lore.kernel.org/r/20200504214443.4485-1-masashi.honma@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agospi: dw: Fix native CS being unset
Sasha Levin [Fri, 12 Jun 2020 00:17:21 +0000 (20:17 -0400)]
spi: dw: Fix native CS being unset

[ Upstream commit 9aea644ca17b94f82ad7fa767cbc4509642f4420 ]

Commit 6e0a32d6f376 ("spi: dw: Fix default polarity of native
chipselect") attempted to fix the problem when GPIO active-high
chip-select is utilized to communicate with some SPI slave. It fixed
the problem, but broke the normal native CS support. At the same time
the reversion commit ada9e3fcc175 ("spi: dw: Correct handling of native
chipselect") didn't solve the problem either, since it just inverted
the set_cs() polarity perception without taking into account that
CS-high might be applicable. Here is what is done to finally fix the
problem.

DW SPI controller demands any native CS being set in order to proceed
with data transfer. So in order to activate the SPI communications we
must set any bit in the Slave Select DW SPI controller register no
matter whether the platform requests the GPIO- or native CS. Preferably
it should be the bit corresponding to the SPI slave CS number. But
currently the dw_spi_set_cs() method activates the chip-select
only if the second argument is false. Since the second argument of the
set_cs callback is expected to be a boolean with "is-high" semantics
(actual chip-select pin state value), the bit in the DW SPI Slave
Select register will be set only if SPI core requests the driver
to set the CS in the low state. So this will work for active-low
GPIO-based CS case, and won't work for active-high CS setting
the bit when SPI core actually needs to deactivate the CS.

This commit fixes the problem for all described cases. So no matter
whether an SPI slave needs GPIO- or native-based CS with active-high
or low signal the corresponding bit will be set in SER.

Signed-off-by: Serge Semin <Sergey.Semin@baikalelectronics.ru>
Fixes: ada9e3fcc175 ("spi: dw: Correct handling of native chipselect")
Fixes: 6e0a32d6f376 ("spi: dw: Fix default polarity of native chipselect")
Reviewed-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Link: https://lore.kernel.org/r/20200515104758.6934-5-Sergey.Semin@baikalelectronics.ru
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agopowerpc/xive: Clear the page tables for the ESB IO mapping
Cédric Le Goater [Wed, 29 Apr 2020 07:51:20 +0000 (09:51 +0200)]
powerpc/xive: Clear the page tables for the ESB IO mapping

[ Upstream commit a101950fcb78b0ba20cd487be6627dea58d55c2b ]

Commit 1ca3dec2b2df ("powerpc/xive: Prevent page fault issues in the
machine crash handler") fixed an issue in the FW assisted dump of
machines using hash MMU and the XIVE interrupt mode under the POWER
hypervisor. It forced the mapping of the ESB page of interrupts being
mapped in the Linux IRQ number space to make sure the 'crash kexec'
sequence worked during such an event. But it didn't handle the
un-mapping.

This mapping is now blocking the removal of a passthrough IO adapter
under the POWER hypervisor because it expects the guest OS to have
cleared all page table entries related to the adapter. If some are
still present, the RTAS call which isolates the PCI slot returns error
9001 "valid outstanding translations".

Remove these mapping in the IRQ data cleanup routine.

Under KVM, this cleanup is not required because the ESB pages for the
adapter interrupts are un-mapped from the guest by the hypervisor in
the KVM XIVE native device. This is now redundant but it's harmless.

Fixes: 1ca3dec2b2df ("powerpc/xive: Prevent page fault issues in the machine crash handler")
Cc: stable@vger.kernel.org # v5.5+
Signed-off-by: Cédric Le Goater <clg@kaod.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20200429075122.1216388-2-clg@kaod.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agodrivers/net/ibmvnic: Update VNIC protocol version reporting
Thomas Falcon [Thu, 28 May 2020 16:19:17 +0000 (11:19 -0500)]
drivers/net/ibmvnic: Update VNIC protocol version reporting

[ Upstream commit 784688993ebac34dffe44a9f2fabbe126ebfd4db ]

VNIC protocol version is reported in big-endian format, but it
is not byteswapped before logging. Fix that, and remove version
comparison as only one protocol version exists at this time.

Signed-off-by: Thomas Falcon <tlfalcon@linux.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agogfs2: Even more gfs2_find_jhead fixes
Andreas Gruenbacher [Tue, 26 May 2020 18:11:51 +0000 (20:11 +0200)]
gfs2: Even more gfs2_find_jhead fixes

[ Upstream commit 20be493b787cd581c9fffad7fcd6bfbe6af1050c ]

Fix several issues in the previous gfs2_find_jhead fix:
* When updating @blocks_submitted, @block refers to the first block block not
  submitted yet, not the last block submitted, so fix an off-by-one error.
* We want to ensure that @blocks_submitted is far enough ahead of @blocks_read
  to guarantee that there is in-flight I/O.  Otherwise, we'll eventually end up
  waiting for pages that haven't been submitted, yet.
* It's much easier to compare the number of blocks added with the number of
  blocks submitted to limit the maximum bio size.
* Even with bio chaining, we can keep adding blocks until we reach the maximum
  bio size, as long as we stop at a page boundary.  This simplifies the logic.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Reviewed-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agocsky: Fixup abiv2 syscall_trace break a4 & a5
Guo Ren [Sun, 24 May 2020 10:44:38 +0000 (10:44 +0000)]
csky: Fixup abiv2 syscall_trace break a4 & a5

[ Upstream commit e0bbb53843b5fdfe464b099217e3b9d97e8a75d7 ]

Current implementation could destory a4 & a5 when strace, so we need to get them
from pt_regs by SAVE_ALL.

Signed-off-by: Guo Ren <guoren@linux.alibaba.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoInput: synaptics - add a second working PNP_ID for Lenovo T470s
Dennis Kadioglu [Wed, 27 May 2020 06:03:13 +0000 (23:03 -0700)]
Input: synaptics - add a second working PNP_ID for Lenovo T470s

[ Upstream commit 642aa86eaf8f1e6fe894f20fd7f12f0db52ee03c ]

The Lenovo Thinkpad T470s I own has a different touchpad with "LEN007a"
instead of the already included PNP ID "LEN006c". However, my touchpad
seems to work well without any problems using RMI. So this patch adds the
other PNP ID.

Signed-off-by: Dennis Kadioglu <denk@eclipso.email>
Link: https://lore.kernel.org/r/ff770543cd53ae818363c0fe86477965@mail.eclipso.de
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agosched/fair: Don't NUMA balance for kthreads
Jens Axboe [Tue, 26 May 2020 15:38:31 +0000 (09:38 -0600)]
sched/fair: Don't NUMA balance for kthreads

[ Upstream commit 18f855e574d9799a0e7489f8ae6fd8447d0dd74a ]

Stefano reported a crash with using SQPOLL with io_uring:

  BUG: kernel NULL pointer dereference, address: 00000000000003b0
  CPU: 2 PID: 1307 Comm: io_uring-sq Not tainted 5.7.0-rc7 #11
  RIP: 0010:task_numa_work+0x4f/0x2c0
  Call Trace:
   task_work_run+0x68/0xa0
   io_sq_thread+0x252/0x3d0
   kthread+0xf9/0x130
   ret_from_fork+0x35/0x40

which is task_numa_work() oopsing on current->mm being NULL.

The task work is queued by task_tick_numa(), which checks if current->mm is
NULL at the time of the call. But this state isn't necessarily persistent,
if the kthread is using use_mm() to temporarily adopt the mm of a task.

Change the task_tick_numa() check to exclude kernel threads in general,
as it doesn't make sense to attempt ot balance for kthreads anyway.

Reported-by: Stefano Garzarella <sgarzare@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/865de121-8190-5d30-ece5-3b097dc74431@kernel.dk
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoARM: 8977/1: ptrace: Fix mask for thumb breakpoint hook
Fredrik Strupe [Mon, 18 May 2020 18:41:11 +0000 (19:41 +0100)]
ARM: 8977/1: ptrace: Fix mask for thumb breakpoint hook

[ Upstream commit 3866f217aaa81bf7165c7f27362eee5d7919c496 ]

call_undef_hook() in traps.c applies the same instr_mask for both 16-bit
and 32-bit thumb instructions. If instr_mask then is only 16 bits wide
(0xffff as opposed to 0xffffffff), the first half-word of 32-bit thumb
instructions will be masked out. This makes the function match 32-bit
thumb instructions where the second half-word is equal to instr_val,
regardless of the first half-word.

The result in this case is that all undefined 32-bit thumb instructions
with the second half-word equal to 0xde01 (udf #1) work as breakpoints
and will raise a SIGTRAP instead of a SIGILL, instead of just the one
intended 16-bit instruction. An example of such an instruction is
0xeaa0de01, which is unallocated according to Arm ARM and should raise a
SIGILL, but instead raises a SIGTRAP.

This patch fixes the issue by setting all the bits in instr_mask, which
will still match the intended 16-bit thumb instruction (where the
upper half is always 0), but not any 32-bit thumb instructions.

Cc: Oleg Nesterov <oleg@redhat.com>
Signed-off-by: Fredrik Strupe <fredrik@strupe.net>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoInput: mms114 - fix handling of mms345l
Stephan Gerhold [Thu, 9 Apr 2020 20:49:03 +0000 (13:49 -0700)]
Input: mms114 - fix handling of mms345l

[ Upstream commit 3f8f770575d911c989043d8f0fb8dec96360c41c ]

MMS345L is another first generation touch screen from Melfas,
which uses the same registers as MMS152.

However, using I2C_M_NOSTART for it causes errors when reading:

i2c i2c-0: sendbytes: NAK bailout.
mms114 0-0048: __mms114_read_reg: i2c transfer failed (-5)

The driver works fine as soon as I2C_M_NOSTART is removed.

Reviewed-by: Andi Shyti <andi@etezian.org>
Signed-off-by: Stephan Gerhold <stephan@gerhold.net>
Link: https://lore.kernel.org/r/20200405170904.61512-1-stephan@gerhold.net
[dtor: removed separate mms345l handling, made everyone use standard
transfer mode, propagated the 10bit addressing flag to the read part of the
transfer as well.]
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoelfnote: mark all .note sections SHF_ALLOC
Nick Desaulniers [Thu, 4 Jun 2020 23:50:49 +0000 (16:50 -0700)]
elfnote: mark all .note sections SHF_ALLOC

commit 51da9dfb7f20911ae4e79e9b412a9c2d4c373d4b upstream.

ELFNOTE_START allows callers to specify flags for .pushsection assembler
directives.  All callsites but ELF_NOTE use "a" for SHF_ALLOC.  For vdso's
that explicitly use ELF_NOTE_START and BUILD_SALT, the same section is
specified twice after preprocessing, once with "a" flag, once without.
Example:

.pushsection .note.Linux, "a", @note ;
.pushsection .note.Linux, "", @note ;

While GNU as allows this ordering, it warns for the opposite ordering,
making these directives position dependent.  We'd prefer not to precisely
match this behavior in Clang's integrated assembler.  Instead, the non
__ASSEMBLY__ definition of ELF_NOTE uses
__attribute__((section(".note.Linux"))) which is created with SHF_ALLOC,
so let's make the __ASSEMBLY__ definition of ELF_NOTE consistent with C
and just always use "a" flag.

This allows Clang to assemble a working mainline (5.6) kernel via:
$ make CC=clang AS=clang

Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: Fangrui Song <maskray@google.com>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Vincenzo Frascino <vincenzo.frascino@arm.com>
Link: https://github.com/ClangBuiltLinux/linux/issues/913
Link: http://lkml.kernel.org/r/20200325231250.99205-1-ndesaulniers@google.com
Debugged-by: Ilie Halip <ilie.halip@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Jian Cai <jiancai@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobpf: Support llvm-objcopy for vmlinux BTF
Fangrui Song [Wed, 18 Mar 2020 22:27:46 +0000 (15:27 -0700)]
bpf: Support llvm-objcopy for vmlinux BTF

commit 90ceddcb495008ac8ba7a3dce297841efcd7d584 upstream.

Simplify gen_btf logic to make it work with llvm-objcopy. The existing
'file format' and 'architecture' parsing logic is brittle and does not
work with llvm-objcopy/llvm-objdump.

'file format' output of llvm-objdump>=11 will match GNU objdump, but
'architecture' (bfdarch) may not.

.BTF in .tmp_vmlinux.btf is non-SHF_ALLOC. Add the SHF_ALLOC flag
because it is part of vmlinux image used for introspection. C code
can reference the section via linker script defined __start_BTF and
__stop_BTF. This fixes a small problem that previous .BTF had the
SHF_WRITE flag (objcopy -I binary -O elf* synthesized .data).

Additionally, `objcopy -I binary` synthesized symbols
_binary__btf_vmlinux_bin_start and _binary__btf_vmlinux_bin_stop (not
used elsewhere) are replaced with more commonplace __start_BTF and
__stop_BTF.

Add 2>/dev/null because GNU objcopy (but not llvm-objcopy) warns
"empty loadable segment detected at vaddr=0xffffffff81000000, is this intentional?"

We use a dd command to change the e_type field in the ELF header from
ET_EXEC to ET_REL so that lld will accept .btf.vmlinux.bin.o.  Accepting
ET_EXEC as an input file is an extremely rare GNU ld feature that lld
does not intend to support, because this is error-prone.

The output section description .BTF in include/asm-generic/vmlinux.lds.h
avoids potential subtle orphan section placement issues and suppresses
--orphan-handling=warn warnings.

Fixes: df786c9b9476 ("bpf: Force .BTF section start to zero when dumping from vmlinux")
Fixes: cb0cc635c7a9 ("powerpc: Include .BTF section")
Reported-by: Nathan Chancellor <natechancellor@gmail.com>
Signed-off-by: Fangrui Song <maskray@google.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Tested-by: Stanislav Fomichev <sdf@google.com>
Tested-by: Andrii Nakryiko <andriin@fb.com>
Reviewed-by: Stanislav Fomichev <sdf@google.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Acked-by: Andrii Nakryiko <andriin@fb.com>
Acked-by: Michael Ellerman <mpe@ellerman.id.au> (powerpc)
Link: https://github.com/ClangBuiltLinux/linux/issues/871
Link: https://lore.kernel.org/bpf/20200318222746.173648-1-maskray@google.com
Signed-off-by: Maria Teguiani <teguiani@google.com>
Tested-by: Matthias Maennich <maennich@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agovxlan: Avoid infinite loop when suppressing NS messages with invalid options
Ido Schimmel [Mon, 1 Jun 2020 12:58:55 +0000 (15:58 +0300)]
vxlan: Avoid infinite loop when suppressing NS messages with invalid options

[ Upstream commit 8066e6b449e050675df48e7c4b16c29f00507ff0 ]

When proxy mode is enabled the vxlan device might reply to Neighbor
Solicitation (NS) messages on behalf of remote hosts.

In case the NS message includes the "Source link-layer address" option
[1], the vxlan device will use the specified address as the link-layer
destination address in its reply.

To avoid an infinite loop, break out of the options parsing loop when
encountering an option with length zero and disregard the NS message.

This is consistent with the IPv6 ndisc code and RFC 4886 which states
that "Nodes MUST silently discard an ND packet that contains an option
with length zero" [2].

[1] https://tools.ietf.org/html/rfc4861#section-4.3
[2] https://tools.ietf.org/html/rfc4861#section-4.6

Fixes: 4b29dba9c085 ("vxlan: fix nonfunctional neigh_reduce()")
Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Nikolay Aleksandrov <nikolay@cumulusnetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>