ext4: fix race when reusing xattr blocks

[ Upstream commit 65f8b80053a1b2fd602daa6814e62d6fa90e5e9b ]

When ext4_xattr_block_set() decides to remove xattr block the following
race can happen:

CPU1                                    CPU2
ext4_xattr_block_set()                  ext4_xattr_release_block()
  new_bh = ext4_xattr_block_cache_find()

                                          lock_buffer(bh);
                                          ref = le32_to_cpu(BHDR(bh)->h_refcount);
                                          if (ref == 1) {
                                            ...
                                            mb_cache_entry_delete();
                                            unlock_buffer(bh);
                                            ext4_free_blocks();
                                              ...
                                              ext4_forget(..., bh, ...);
                                                jbd2_journal_revoke(..., bh);

  ext4_journal_get_write_access(..., new_bh, ...)
    do_get_write_access()
      jbd2_journal_cancel_revoke(..., new_bh);

Later the code in ext4_xattr_block_set() finds out the block got freed
and cancels reusal of the block but the revoke stays canceled and so in
case of block reuse and journal replay the filesystem can get corrupted.
If the race works out slightly differently, we can also hit assertions
in the jbd2 code.

Fix the problem by making sure that once matching mbcache entry is
found, code dropping the last xattr block reference (or trying to modify
xattr block in place) waits until the mbcache entry reference is
dropped. This way code trying to reuse xattr block is protected from
someone trying to drop the last reference to xattr block.

Reported-and-tested-by: Ritesh Harjani <ritesh.list@gmail.com>
CC: stable@vger.kernel.org
Fixes: 82939d7999df ("ext4: convert to mbcache2")
Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220712105436.32204-5-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: unindent codeblock in ext4_xattr_block_set()

[ Upstream commit fd48e9acdf26d0cbd80051de07d4a735d05d29b2 ]

Remove unnecessary else (and thus indentation level) from a code block
in ext4_xattr_block_set(). It will also make following code changes
easier. No functional changes.

CC: stable@vger.kernel.org
Fixes: 82939d7999df ("ext4: convert to mbcache2")
Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220712105436.32204-4-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: use kmemdup() to replace kmalloc + memcpy

[ Upstream commit 4efd9f0d120c55b08852ee5605dbb02a77089a5d ]

Replace kmalloc + memcpy with kmemdup()

Signed-off-by: Shuqi Zhang <zhangshuqi3@huawei.com>
Reviewed-by: Ritesh Harjani <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20220525030120.803330-1-zhangshuqi3@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: remove EA inode entry from mbcache on inode eviction

[ Upstream commit 6bc0d63dad7f9f54d381925ee855b402f652fa39 ]

Currently we remove EA inode from mbcache as soon as its xattr refcount
drops to zero. However there can be pending attempts to reuse the inode
and thus refcount handling code has to handle the situation when
refcount increases from zero anyway. So save some work and just keep EA
inode in mbcache until it is getting evicted. At that moment we are sure
following iget() of EA inode will fail anyway (or wait for eviction to
finish and load things from the disk again) and so removing mbcache
entry at that moment is fine and simplifies the code a bit.

CC: stable@vger.kernel.org
Fixes: 82939d7999df ("ext4: convert to mbcache2")
Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220712105436.32204-3-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: make sure ext4_append() always allocates new block

[ Upstream commit b8a04fe77ef1360fbf73c80fddbdfeaa9407ed1b ]

ext4_append() must always allocate a new block, otherwise we run the
risk of overwriting existing directory block corrupting the directory
tree in the process resulting in all manner of problems later on.

Add a sanity check to see if the logical block is already allocated and
error out if it is.

Cc: stable@kernel.org
Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20220704142721.157985-2-lczerner@redhat.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: check if directory block is within i_size

[ Upstream commit 65f8ea4cd57dbd46ea13b41dc8bac03176b04233 ]

Currently ext4 directory handling code implicitly assumes that the
directory blocks are always within the i_size. In fact ext4_append()
will attempt to allocate next directory block based solely on i_size and
the i_size is then appropriately increased after a successful
allocation.

However, for this to work it requires i_size to be correct. If, for any
reason, the directory inode i_size is corrupted in a way that the
directory tree refers to a valid directory block past i_size, we could
end up corrupting parts of the directory tree structure by overwriting
already used directory blocks when modifying the directory.

Fix it by catching the corruption early in __ext4_read_dirblock().

Addresses Red-Hat-Bugzilla: #2070205
CVE: CVE-2022-1184
Signed-off-by: Lukas Czerner <lczerner@redhat.com>
Cc: stable@vger.kernel.org
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20220704142721.157985-1-lczerner@redhat.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: fix warning in ext4_iomap_begin as race between bmap and write

[ Upstream commit 51ae846cff568c8c29921b1b28eb2dfbcd4ac12d ]

We got issue as follows:
------------[ cut here ]------------
WARNING: CPU: 3 PID: 9310 at fs/ext4/inode.c:3441 ext4_iomap_begin+0x182/0x5d0
RIP: 0010:ext4_iomap_begin+0x182/0x5d0
RSP: 0018:ffff88812460fa08 EFLAGS: 00010293
RAX: ffff88811f168000 RBX: 0000000000000000 RCX: ffffffff97793c12
RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000003
RBP: ffff88812c669160 R08: ffff88811f168000 R09: ffffed10258cd20f
R10: ffff88812c669077 R11: ffffed10258cd20e R12: 0000000000000001
R13: 00000000000000a4 R14: 000000000000000c R15: ffff88812c6691ee
FS:  00007fd0d6ff3740(0000) GS:ffff8883af180000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007fd0d6dda290 CR3: 0000000104a62000 CR4: 00000000000006e0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 iomap_apply+0x119/0x570
 iomap_bmap+0x124/0x150
 ext4_bmap+0x14f/0x250
 bmap+0x55/0x80
 do_vfs_ioctl+0x952/0xbd0
 __x64_sys_ioctl+0xc6/0x170
 do_syscall_64+0x33/0x40
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

Above issue may happen as follows:
          bmap                    write
bmap
  ext4_bmap
    iomap_bmap
      ext4_iomap_begin
                            ext4_file_write_iter
      ext4_buffered_write_iter
        generic_perform_write
  ext4_da_write_begin
    ext4_da_write_inline_data_begin
      ext4_prepare_inline_data
        ext4_create_inline_data
  ext4_set_inode_flag(inode,
EXT4_INODE_INLINE_DATA);
      if (WARN_ON_ONCE(ext4_has_inline_data(inode))) ->trigger bug_on

To solved above issue hold inode lock in ext4_bamp.

Signed-off-by: Ye Bin <yebin10@huawei.com>
Link: https://lore.kernel.org/r/20220617013935.397596-1-yebin10@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: correct the misjudgment in ext4_iget_extra_inode

[ Upstream commit fd7e672ea98b95b9d4c9dae316639f03c16a749d ]

Use the EXT4_INODE_HAS_XATTR_SPACE macro to more accurately
determine whether the inode have xattr space.

Cc: stable@kernel.org
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220616021358.2504451-5-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: correct max_inline_xattr_value_size computing

[ Upstream commit c9fd167d57133c5b748d16913c4eabc55e531c73 ]

If the ext4 inode does not have xattr space, 0 is returned in the
get_max_inline_xattr_value_size function. Otherwise, the function returns
a negative value when the inode does not contain EXT4_STATE_XATTR.

Cc: stable@kernel.org
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220616021358.2504451-4-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: fix use-after-free in ext4_xattr_set_entry

[ Upstream commit 67d7d8ad99beccd9fe92d585b87f1760dc9018e3 ]

Hulk Robot reported a issue:
==================================================================
BUG: KASAN: use-after-free in ext4_xattr_set_entry+0x18ab/0x3500
Write of size 4105 at addr ffff8881675ef5f4 by task syz-executor.0/7092

CPU: 1 PID: 7092 Comm: syz-executor.0 Not tainted 4.19.90-dirty #17
Call Trace:
[...]
 memcpy+0x34/0x50 mm/kasan/kasan.c:303
 ext4_xattr_set_entry+0x18ab/0x3500 fs/ext4/xattr.c:1747
 ext4_xattr_ibody_inline_set+0x86/0x2a0 fs/ext4/xattr.c:2205
 ext4_xattr_set_handle+0x940/0x1300 fs/ext4/xattr.c:2386
 ext4_xattr_set+0x1da/0x300 fs/ext4/xattr.c:2498
 __vfs_setxattr+0x112/0x170 fs/xattr.c:149
 __vfs_setxattr_noperm+0x11b/0x2a0 fs/xattr.c:180
 __vfs_setxattr_locked+0x17b/0x250 fs/xattr.c:238
 vfs_setxattr+0xed/0x270 fs/xattr.c:255
 setxattr+0x235/0x330 fs/xattr.c:520
 path_setxattr+0x176/0x190 fs/xattr.c:539
 __do_sys_lsetxattr fs/xattr.c:561 [inline]
 __se_sys_lsetxattr fs/xattr.c:557 [inline]
 __x64_sys_lsetxattr+0xc2/0x160 fs/xattr.c:557
 do_syscall_64+0xdf/0x530 arch/x86/entry/common.c:298
 entry_SYSCALL_64_after_hwframe+0x44/0xa9
RIP: 0033:0x459fe9
RSP: 002b:00007fa5e54b4c08 EFLAGS: 00000246 ORIG_RAX: 00000000000000bd
RAX: ffffffffffffffda RBX: 000000000051bf60 RCX: 0000000000459fe9
RDX: 00000000200003c0 RSI: 0000000020000180 RDI: 0000000020000140
RBP: 000000000051bf60 R08: 0000000000000001 R09: 0000000000000000
R10: 0000000000001009 R11: 0000000000000246 R12: 0000000000000000
R13: 00007ffc73c93fc0 R14: 000000000051bf60 R15: 00007fa5e54b4d80
[...]
==================================================================

Above issue may happen as follows:
-------------------------------------
ext4_xattr_set
  ext4_xattr_set_handle
    ext4_xattr_ibody_find
      >> s->end < s->base
      >> no EXT4_STATE_XATTR
      >> xattr_check_inode is not executed
    ext4_xattr_ibody_set
      ext4_xattr_set_entry
       >> size_t min_offs = s->end - s->base
       >> UAF in memcpy

we can easily reproduce this problem with the following commands:
    mkfs.ext4 -F /dev/sda
    mount -o debug_want_extra_isize=128 /dev/sda /mnt
    touch /mnt/file
    setfattr -n user.cat -v `seq -s z 4096|tr -d '[:digit:]'` /mnt/file

In ext4_xattr_ibody_find, we have the following assignment logic:
  header = IHDR(inode, raw_inode)
         = raw_inode + EXT4_GOOD_OLD_INODE_SIZE + i_extra_isize
  is->s.base = IFIRST(header)
             = header + sizeof(struct ext4_xattr_ibody_header)
  is->s.end = raw_inode + s_inode_size

In ext4_xattr_set_entry
  min_offs = s->end - s->base
           = s_inode_size - EXT4_GOOD_OLD_INODE_SIZE - i_extra_isize -
     sizeof(struct ext4_xattr_ibody_header)
  last = s->first
  free = min_offs - ((void *)last - s->base) - sizeof(__u32)
       = s_inode_size - EXT4_GOOD_OLD_INODE_SIZE - i_extra_isize -
         sizeof(struct ext4_xattr_ibody_header) - sizeof(__u32)

In the calculation formula, all values except s_inode_size and
i_extra_size are fixed values. When i_extra_size is the maximum value
s_inode_size - EXT4_GOOD_OLD_INODE_SIZE, min_offs is -4 and free is -8.
The value overflows. As a result, the preceding issue is triggered when
memcpy is executed.

Therefore, when finding xattr or setting xattr, check whether
there is space for storing xattr in the inode to resolve this issue.

Cc: stable@kernel.org
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20220616021358.2504451-3-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: add EXT4_INODE_HAS_XATTR_SPACE macro in xattr.h

[ Upstream commit 179b14152dcb6a24c3415200603aebca70ff13af ]

When adding an xattr to an inode, we must ensure that the inode_size is
not less than EXT4_GOOD_OLD_INODE_SIZE + extra_isize + pad. Otherwise,
the end position may be greater than the start position, resulting in UAF.

Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20220616021358.2504451-2-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: fix extent status tree race in writeback error recovery path

[ Upstream commit 7f0d8e1d607c1a4fa9a27362a108921d82230874 ]

A race can occur in the unlikely event ext4 is unable to allocate a
physical cluster for a delayed allocation in a bigalloc file system
during writeback.  Failure to allocate a cluster forces error recovery
that includes a call to mpage_release_unused_pages().  That function
removes any corresponding delayed allocated blocks from the extent
status tree.  If a new delayed write is in progress on the same cluster
simultaneously, resulting in the addition of an new extent containing
one or more blocks in that cluster to the extent status tree, delayed
block accounting can be thrown off if that delayed write then encounters
a similar cluster allocation failure during future writeback.

Write lock the i_data_sem in mpage_release_unused_pages() to fix this
problem.  Ext4's block/cluster accounting code for bigalloc relies on
i_data_sem for mutual exclusion, as is found in the delayed write path,
and the locking in mpage_release_unused_pages() is missing.

Cc: stable@kernel.org
Reported-by: Ye Bin <yebin10@huawei.com>
Signed-off-by: Eric Whitney <enwlinux@gmail.com>
Link: https://lore.kernel.org/r/20220615160530.1928801-1-enwlinux@gmail.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ext4: update s_overhead_clusters in the superblock during an on-line resize

[ Upstream commit de394a86658ffe4e89e5328fd4993abfe41b7435 ]

When doing an online resize, the on-disk superblock on-disk wasn't
updated.  This means that when the file system is unmounted and
remounted, and the on-disk overhead value is non-zero, this would
result in the results of statfs(2) to be incorrect.

This was partially fixed by Commits 10b01ee92df5 ("ext4: fix overhead
calculation to account for the reserved gdt blocks"), 85d825dbf489
("ext4: force overhead calculation if the s_overhead_cluster makes no
sense"), and eb7054212eac ("ext4: update the cached overhead value in
the superblock").

However, since it was too expensive to forcibly recalculate the
overhead for bigalloc file systems at every mount, this didn't fix the
problem for bigalloc file systems.  This commit should address the
problem when resizing file systems with the bigalloc feature enabled.

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Reviewed-by: Andreas Dilger <adilger@dilger.ca>
Link: https://lore.kernel.org/r/20220629040026.112371-1-tytso@mit.edu
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tracing: Avoid -Warray-bounds warning for __rel_loc macro

[ Upstream commit 58c5724ec2cdd72b22107ec5de00d90cc4797796 ]

Since -Warray-bounds checks the destination size from the type of given
pointer, __assign_rel_str() macro gets warned because it passes the
pointer to the 'u32' field instead of 'trace_event_raw_*' data structure.
Pass the data address calculated from the 'trace_event_raw_*' instead of
'u32' __rel_loc field.

Link: https://lkml.kernel.org/r/20220125233154.dac280ed36944c0c2fe6f3ac@kernel.org
Cc: Stephen Rothwell <sfr@canb.auug.org.au>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
[ This did not fix the warning, but is still a nice clean up ]
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tracing: Add '__rel_loc' using trace event macros

[ Upstream commit 55de2c0b5610cba5a5a93c0788031133c457e689 ]

Add '__rel_loc' using trace event macros. These macros are usually
not used in the kernel, except for testing purpose.
This also add "rel_" variant of macros for dynamic_array string,
and bitmask.

Link: https://lkml.kernel.org/r/163757342119.510314.816029622439099016.stgit@devnote2
Cc: Beau Belgrave <beaub@linux.microsoft.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Tom Zanussi <zanussi@kernel.org>
Signed-off-by: Masami Hiramatsu <mhiramat@kernel.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dm raid: fix address sanitizer warning in raid_resume

[ Upstream commit 7dad24db59d2d2803576f2e3645728866a056dab ]

There is a KASAN warning in raid_resume when running the lvm test
lvconvert-raid.sh. The reason for the warning is that mddev->raid_disks
is greater than rs->raid_disks, so the loop touches one entry beyond
the allocated length.

Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dm raid: fix address sanitizer warning in raid_status

[ Upstream commit 1fbeea217d8f297fe0e0956a1516d14ba97d0396 ]

There is this warning when using a kernel with the address sanitizer
and running this testsuite:
https://gitlab.com/cki-project/kernel-tests/-/tree/main/storage/swraid/scsi_raid

==================================================================
BUG: KASAN: slab-out-of-bounds in raid_status+0x1747/0x2820 [dm_raid]
Read of size 4 at addr ffff888079d2c7e8 by task lvcreate/13319
CPU: 0 PID: 13319 Comm: lvcreate Not tainted 5.18.0-0.rc3.<snip> #1
Hardware name: Red Hat KVM, BIOS 0.5.1 01/01/2011
Call Trace:
 <TASK>
 dump_stack_lvl+0x6a/0x9c
 print_address_description.constprop.0+0x1f/0x1e0
 print_report.cold+0x55/0x244
 kasan_report+0xc9/0x100
 raid_status+0x1747/0x2820 [dm_raid]
 dm_ima_measure_on_table_load+0x4b8/0xca0 [dm_mod]
 table_load+0x35c/0x630 [dm_mod]
 ctl_ioctl+0x411/0x630 [dm_mod]
 dm_ctl_ioctl+0xa/0x10 [dm_mod]
 __x64_sys_ioctl+0x12a/0x1a0
 do_syscall_64+0x5b/0x80

The warning is caused by reading conf->max_nr_stripes in raid_status. The
code in raid_status reads mddev->private, casts it to struct r5conf and
reads the entry max_nr_stripes.

However, if we have different raid type than 4/5/6, mddev->private
doesn't point to struct r5conf; it may point to struct r0conf, struct
r1conf, struct r10conf or struct mpconf. If we cast a pointer to one
of these structs to struct r5conf, we will be reading invalid memory
and KASAN warns about it.

Fix this bug by reading struct r5conf only if raid type is 4, 5 or 6.

Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

KVM: nVMX: Attempt to load PERF_GLOBAL_CTRL on nVMX xfer iff it exists

[ Upstream commit 4496a6f9b45e8cd83343ad86a3984d614e22cf54 ]

Attempt to load PERF_GLOBAL_CTRL during nested VM-Enter/VM-Exit if and
only if the MSR exists (according to the guest vCPU model).  KVM has very
misguided handling of VM_{ENTRY,EXIT}_LOAD_IA32_PERF_GLOBAL_CTRL and
attempts to force the nVMX MSR settings to match the vPMU model, i.e. to
hide/expose the control based on whether or not the MSR exists from the
guest's perspective.

KVM's modifications fail to handle the scenario where the vPMU is hidden
from the guest _after_ being exposed to the guest, e.g. by userspace
doing multiple KVM_SET_CPUID2 calls, which is allowed if done before any
KVM_RUN.  nested_vmx_pmu_refresh() is called if and only if there's a
recognized vPMU, i.e. KVM will leave the bits in the allow state and then
ultimately reject the MSR load and WARN.

KVM should not force the VMX MSRs in the first place.  KVM taking control
of the MSRs was a misguided attempt at mimicking what commit 5f76f6f5ff96
("KVM: nVMX: Do not expose MPX VMX controls when guest MPX disabled",
2018-10-01) did for MPX.  However, the MPX commit was a workaround for
another KVM bug and not something that should be imitated (and it should
never been done in the first place).

In other words, KVM's ABI _should_ be that userspace has full control
over the MSRs, at which point triggering the WARN that loading the MSR
must not fail is trivial.

The intent of the WARN is still valid; KVM has consistency checks to
ensure that vmcs12->{guest,host}_ia32_perf_global_ctrl is valid.  The
problem is that '0' must be considered a valid value at all times, and so
the simple/obvious solution is to just not actually load the MSR when it
does not exist.  It is userspace's responsibility to provide a sane vCPU
model, i.e. KVM is well within its ABI and Intel's VMX architecture to
skip the loads if the MSR does not exist.

Fixes: 03a8871add95 ("KVM: nVMX: Expose load IA32_PERF_GLOBAL_CTRL VM-{Entry,Exit} control")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220722224409.1336532-5-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

KVM: VMX: Add helper to check if the guest PMU has PERF_GLOBAL_CTRL

[ Upstream commit b663f0b5f3d665c261256d1f76e98f077c6e56af ]

Add a helper to check of the guest PMU has PERF_GLOBAL_CTRL, which is
unintuitive _and_ diverges from Intel's architecturally defined behavior.
Even worse, KVM currently implements the check using two different (but
equivalent) checks, _and_ there has been at least one attempt to add a
_third_ flavor.

Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220722224409.1336532-4-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

KVM: x86/pmu: Ignore pmu->global_ctrl check if vPMU doesn't support global_ctrl

[ Upstream commit 98defd2e17803263f49548fea930cfc974d505aa ]

MSR_CORE_PERF_GLOBAL_CTRL is introduced as part of Architecture PMU V2,
as indicated by Intel SDM 19.2.2 and the intel_is_valid_msr() function.

So in the absence of global_ctrl support, all PMCs are enabled as AMD does.

Signed-off-by: Like Xu <likexu@tencent.com>
Message-Id: <20220509102204.62389-1-likexu@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

KVM: VMX: Mark all PERF_GLOBAL_(OVF)_CTRL bits reserved if there's no vPMU

[ Upstream commit 93255bf92939d948bc86d81c6bb70bb0fecc5db1 ]

Mark all MSR_CORE_PERF_GLOBAL_CTRL and MSR_CORE_PERF_GLOBAL_OVF_CTRL bits
as reserved if there is no guest vPMU.  The nVMX VM-Entry consistency
checks do not check for a valid vPMU prior to consuming the masks via
kvm_valid_perf_global_ctrl(), i.e. may incorrectly allow a non-zero mask
to be loaded via VM-Enter or VM-Exit (well, attempted to be loaded, the
actual MSR load will be rejected by intel_is_valid_msr()).

Fixes: f5132b01386b ("KVM: Expose a version 2 architectural PMU to a guests")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Message-Id: <20220722224409.1336532-3-seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

KVM: x86/pmu: Introduce the ctrl_mask value for fixed counter

[ Upstream commit 2c985527dd8d283e786ad7a67e532ef7f6f00fac ]

The mask value of fixed counter control register should be dynamic
adjusted with the number of fixed counters. This patch introduces a
variable that includes the reserved bits of fixed counter control
registers. This is a generic code refactoring.

Co-developed-by: Luwei Kang <luwei.kang@intel.com>
Signed-off-by: Luwei Kang <luwei.kang@intel.com>
Signed-off-by: Like Xu <like.xu@linux.intel.com>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Message-Id: <20220411101946.20262-6-likexu@tencent.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc/powernv/kvm: Use darn for H_RANDOM on Power9

[ Upstream commit 7ef3d06f1bc4a5e62273726f3dc2bd258ae1c71f ]

The existing logic in KVM to support guests calling H_RANDOM only works
on Power8, because it looks for an RNG in the device tree, but on Power9
we just use darn.

In addition the existing code needs to work in real mode, so we have the
special cased powernv_get_random_real_mode() to deal with that.

Instead just have KVM call ppc_md.get_random_seed(), and do the real
mode check inside of there, that way we use whatever RNG is available,
including darn on Power9.

Fixes: e928e9cb3601 ("KVM: PPC: Book3S HV: Add fast real-mode H_RANDOM implementation.")
Cc: stable@vger.kernel.org # v4.1+
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Tested-by: Sachin Sant <sachinp@linux.ibm.com>
[mpe: Rebase on previous commit, update change log appropriately]
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20220727143219.2684192-2-mpe@ellerman.id.au
Signed-off-by: Sasha Levin <sashal@kernel.org>

ACPI: CPPC: Do not prevent CPPC from working in the future

[ Upstream commit 4f4179fcf420873002035cf1941d844c9e0e7cb3 ]

There is a problem with the current revision checks in
is_cppc_supported() that they essentially prevent the CPPC support
from working if a new _CPC package format revision being a proper
superset of the v3 and only causing _CPC to return a package with more
entries (while retaining the types and meaning of the entries defined by
the v3) is introduced in the future and used by the platform firmware.

In that case, as long as the number of entries in the _CPC return
package is at least CPPC_V3_NUM_ENT, it should be perfectly fine to
use the v3 support code and disregard the additional package entries
added by the new package format revision.

For this reason, drop is_cppc_supported() altogether, put the revision
checks directly into acpi_cppc_processor_probe() so they are easier to
follow and rework them to take the case mentioned above into account.

Fixes: 4773e77cdc9b ("ACPI / CPPC: Add support for CPPC v3")
Cc: 4.18+ <stable@vger.kernel.org> # 4.18+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

btrfs: properly flag filesystem with BTRFS_FEATURE_INCOMPAT_BIG_METADATA

[ Upstream commit e26b04c4c91925dba57324db177a24e18e2d0013 ]

Commit 6f93e834fa7c seemingly inadvertently moved the code responsible
for flagging the filesystem as having BIG_METADATA to a place where
setting the flag was essentially lost. This means that
filesystems created with kernels containing this bug (starting with 5.15)
can potentially be mounted by older (pre-3.4) kernels. In reality
chances for this happening are low because there are other incompat
flags introduced in the mean time. Still the correct behavior is to set
INCOMPAT_BIG_METADATA flag and persist this in the superblock.

Fixes: 6f93e834fa7c ("btrfs: fix upper limit for max_inline for page size 64K")
CC: stable@vger.kernel.org # 5.4+
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

btrfs: reset block group chunk force if we have to wait

[ Upstream commit 1314ca78b2c35d3e7d0f097268a2ee6dc0d369ef ]

If you try to force a chunk allocation, but you race with another chunk
allocation, you will end up waiting on the chunk allocation that just
occurred and then allocate another chunk.  If you have many threads all
doing this at once you can way over-allocate chunks.

Fix this by resetting force to NO_FORCE, that way if we think we need to
allocate we can, otherwise we don't force another chunk allocation if
one is already happening.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
CC: stable@vger.kernel.org # 5.4+
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

btrfs: ensure pages are unlocked on cow_file_range() failure

[ Upstream commit 9ce7466f372d83054c7494f6b3e4b9abaf3f0355 ]

There is a hung_task report on zoned btrfs like below.

https://github.com/naota/linux/issues/59

  [726.328648] INFO: task rocksdb:high0:11085 blocked for more than 241 seconds.
  [726.329839]       Not tainted 5.16.0-rc1+ #1
  [726.330484] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
  [726.331603] task:rocksdb:high0   state:D stack:    0 pid:11085 ppid: 11082 flags:0x00000000
  [726.331608] Call Trace:
  [726.331611]  <TASK>
  [726.331614]  __schedule+0x2e5/0x9d0
  [726.331622]  schedule+0x58/0xd0
  [726.331626]  io_schedule+0x3f/0x70
  [726.331629]  __folio_lock+0x125/0x200
  [726.331634]  ? find_get_entries+0x1bc/0x240
  [726.331638]  ? filemap_invalidate_unlock_two+0x40/0x40
  [726.331642]  truncate_inode_pages_range+0x5b2/0x770
  [726.331649]  truncate_inode_pages_final+0x44/0x50
  [726.331653]  btrfs_evict_inode+0x67/0x480
  [726.331658]  evict+0xd0/0x180
  [726.331661]  iput+0x13f/0x200
  [726.331664]  do_unlinkat+0x1c0/0x2b0
  [726.331668]  __x64_sys_unlink+0x23/0x30
  [726.331670]  do_syscall_64+0x3b/0xc0
  [726.331674]  entry_SYSCALL_64_after_hwframe+0x44/0xae
  [726.331677] RIP: 0033:0x7fb9490a171b
  [726.331681] RSP: 002b:00007fb943ffac68 EFLAGS: 00000246 ORIG_RAX: 0000000000000057
  [726.331684] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007fb9490a171b
  [726.331686] RDX: 00007fb943ffb040 RSI: 000055a6bbe6ec20 RDI: 00007fb94400d300
  [726.331687] RBP: 00007fb943ffad00 R08: 0000000000000000 R09: 0000000000000000
  [726.331688] R10: 0000000000000031 R11: 0000000000000246 R12: 00007fb943ffb000
  [726.331690] R13: 00007fb943ffb040 R14: 0000000000000000 R15: 00007fb943ffd260
  [726.331693]  </TASK>

While we debug the issue, we found running fstests generic/551 on 5GB
non-zoned null_blk device in the emulated zoned mode also had a
similar hung issue.

Also, we can reproduce the same symptom with an error injected
cow_file_range() setup.

The hang occurs when cow_file_range() fails in the middle of
allocation. cow_file_range() called from do_allocation_zoned() can
split the give region ([start, end]) for allocation depending on
current block group usages. When btrfs can allocate bytes for one part
of the split regions but fails for the other region (e.g. because of
-ENOSPC), we return the error leaving the pages in the succeeded regions
locked. Technically, this occurs only when @unlock == 0. Otherwise, we
unlock the pages in an allocated region after creating an ordered
extent.

Considering the callers of cow_file_range(unlock=0) won't write out
the pages, we can unlock the pages on error exit from
cow_file_range(). So, we can ensure all the pages except @locked_page
are unlocked on error case.

In summary, cow_file_range now behaves like this:

- page_started == 1 (return value)
  - All the pages are unlocked. IO is started.
- unlock == 1
  - All the pages except @locked_page are unlocked in any case
- unlock == 0
  - On success, all the pages are locked for writing out them
  - On failure, all the pages except @locked_page are unlocked

Fixes: 42c011000963 ("btrfs: zoned: introduce dedicated data write path for zoned filesystems")
CC: stable@vger.kernel.org # 5.12+
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

block: don't allow the same type rq_qos add more than once

[ Upstream commit 14a6e2eb7df5c7897c15b109cba29ab0c4a791b6 ]

In our test of iocost, we encountered some list add/del corruptions of
inner_walk list in ioc_timer_fn.

The reason can be described as follows:

cpu 0 cpu 1
ioc_qos_write ioc_qos_write

ioc = q_to_ioc(queue);
if (!ioc) {
        ioc = kzalloc();
ioc = q_to_ioc(queue);
if (!ioc) {
ioc = kzalloc();
...
rq_qos_add(q, rqos);
}
        ...
        rq_qos_add(q, rqos);
        ...
}

When the io.cost.qos file is written by two cpus concurrently, rq_qos may
be added to one disk twice. In that case, there will be two iocs enabled
and running on one disk. They own different iocgs on their active list. In
the ioc_timer_fn function, because of the iocgs from two iocs have the
same root iocg, the root iocg's walk_list may be overwritten by each other
and this leads to list add/del corruptions in building or destroying the
inner_walk list.

And so far, the blk-rq-qos framework works in case that one instance for
one type rq_qos per queue by default. This patch make this explicit and
also fix the crash above.

Signed-off-by: Jinke Han <hanjinke.666@bytedance.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Acked-by: Tejun Heo <tj@kernel.org>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220720093616.70584-1-hanjinke.666@bytedance.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

block: remove the struct blk_queue_ctx forward declaration

[ Upstream commit 9778ac77c2027827ffdbb33d3e936b3a0ae9f0f9 ]

This type doesn't exist at all, so no need to forward declare it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20210920123328.1399408-12-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>

locking/csd_lock: Change csdlock_debug from early_param to __setup

[ Upstream commit 9c9b26b0df270d4f9246e483a44686fca951a29c ]

The csdlock_debug kernel-boot parameter is parsed by the
early_param() function csdlock_debug().  If set, csdlock_debug()
invokes static_branch_enable() to enable csd_lock_wait feature, which
triggers a panic on arm64 for kernels built with CONFIG_SPARSEMEM=y and
CONFIG_SPARSEMEM_VMEMMAP=n.

With CONFIG_SPARSEMEM_VMEMMAP=n, __nr_to_section is called in
static_key_enable() and returns NULL, resulting in a NULL dereference
because mem_section is initialized only later in sparse_init().

This is also a problem for powerpc because early_param() functions
are invoked earlier than jump_label_init(), also resulting in
static_key_enable() failures.  These failures cause the warning "static
key 'xxx' used before call to jump_label_init()".

Thus, early_param is too early for csd_lock_wait to run
static_branch_enable(), so changes it to __setup to fix these.

Fixes: 8d0968cc6b8f ("locking/csd_lock: Add boot parameter for controlling CSD lock debugging")
Cc: stable@vger.kernel.org
Reported-by: Chen jingwen <chenjingwen6@huawei.com>
Signed-off-by: Chen Zhongjin <chenzhongjin@huawei.com>
Signed-off-by: Paul E. McKenney <paulmck@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

timekeeping: contribute wall clock to rng on time change

[ Upstream commit b8ac29b40183a6038919768b5d189c9bd91ce9b4 ]

The rng's random_init() function contributes the real time to the rng at
boot time, so that events can at least start in relation to something
particular in the real world. But this clock might not yet be set that
point in boot, so nothing is contributed. In addition, the relation
between minor clock changes from, say, NTP, and the cycle counter is
potentially useful entropic data.

This commit addresses this by mixing in a time stamp on calls to
settimeofday and adjtimex. No entropy is credited in doing so, so it
doesn't make initialization faster, but it is still useful input to
have.

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Cc: stable@vger.kernel.org
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ARM: remove some dead code

[ Upstream commit 08572cd41955166e387d9b4984294d37f8f7526c ]

This code appears to be no longer used so let's get rid of it.

Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Keith Packard <keithpac@amazon.com>
Tested-by: Marc Zyngier <maz@kernel.org>
Tested-by: Vladimir Murzin <vladimir.murzin@arm.com> # ARMv7M
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/9p: Initialize the iounit field during fid creation

[ Upstream commit aa7aeee169480e98cf41d83c01290a37e569be6d ]

Ensure that the fid's iounit field is set to zero when a new fid is
created. Certain 9P operations, such as OPEN and CREATE, allow the
server to reply with an iounit size which the client code assigns to the
p9_fid struct shortly after the fid is created by p9_fid_create(). On
the other hand, an XATTRWALK operation doesn't allow for the server to
specify an iounit value. The iounit field of the newly allocated p9_fid
struct remained uninitialized in that case. Depending on allocation
patterns, the iounit value could have been something reasonable that was
carried over from previously freed fids or, in the worst case, could
have been arbitrary values from non-fid related usages of the memory
location.

The bug was detected in the Windows Subsystem for Linux 2 (WSL2) kernel
after the uninitialized iounit field resulted in the typical sequence of
two getxattr(2) syscalls, one to get the size of an xattr and another
after allocating a sufficiently sized buffer to fit the xattr value, to
hit an unexpected ERANGE error in the second call to getxattr(2). An
uninitialized iounit field would sometimes force rsize to be smaller
than the xattr value size in p9_client_read_once() and the 9P server in
WSL refused to chunk up the READ on the attr_fid and, instead, returned
ERANGE to the client. The virtfs server in QEMU seems happy to chunk up
the READ and this problem goes undetected there.

Link: https://lkml.kernel.org/r/20220710141402.803295-1-tyhicks@linux.microsoft.com
Fixes: ebf46264a004 ("fs/9p: Add support user. xattr")
Cc: stable@vger.kernel.org
Signed-off-by: Tyler Hicks <tyhicks@linux.microsoft.com>
Reviewed-by: Christian Schoenebeck <linux_oss@crudebyte.com>
Signed-off-by: Dominique Martinet <asmadeus@codewreck.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dm thin: fix use-after-free crash in dm_sm_register_threshold_callback

[ Upstream commit 3534e5a5ed2997ca1b00f44a0378a075bd05e8a3 ]

Fault inject on pool metadata device reports:
  BUG: KASAN: use-after-free in dm_pool_register_metadata_threshold+0x40/0x80
  Read of size 8 at addr ffff8881b9d50068 by task dmsetup/950

  CPU: 7 PID: 950 Comm: dmsetup Tainted: G        W         5.19.0-rc6 #1
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.14.0-1.fc33 04/01/2014
  Call Trace:
   <TASK>
   dump_stack_lvl+0x34/0x44
   print_address_description.constprop.0.cold+0xeb/0x3f4
   kasan_report.cold+0xe6/0x147
   dm_pool_register_metadata_threshold+0x40/0x80
   pool_ctr+0xa0a/0x1150
   dm_table_add_target+0x2c8/0x640
   table_load+0x1fd/0x430
   ctl_ioctl+0x2c4/0x5a0
   dm_ctl_ioctl+0xa/0x10
   __x64_sys_ioctl+0xb3/0xd0
   do_syscall_64+0x35/0x80
   entry_SYSCALL_64_after_hwframe+0x46/0xb0

This can be easily reproduced using:
  echo offline > /sys/block/sda/device/state
  dd if=/dev/zero of=/dev/mapper/thin bs=4k count=10
  dmsetup load pool --table "0 20971520 thin-pool /dev/sda /dev/sdb 128 0 0"

If a metadata commit fails, the transaction will be aborted and the
metadata space maps will be destroyed. If a DM table reload then
happens for this failed thin-pool, a use-after-free will occur in
dm_sm_register_threshold_callback (called from
dm_pool_register_metadata_threshold).

Fix this by in dm_pool_register_metadata_threshold() by returning the
-EINVAL error if the thin-pool is in fail mode. Also fail pool_ctr()
with a new error message: "Error registering metadata threshold".

Fixes: ac8c3f3df65e4 ("dm thin: generate event when metadata threshold passed")
Cc: stable@vger.kernel.org
Reported-by: Hulk Robot <hulkci@huawei.com>
Signed-off-by: Luo Meng <luomeng12@huawei.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

kexec, KEYS, s390: Make use of built-in and secondary keyring for signature verification

[ Upstream commit 0828c4a39be57768b8788e8cbd0d84683ea757e5 ]

commit e23a8020ce4e ("s390/kexec_file: Signature verification prototype")
adds support for KEXEC_SIG verification with keys from platform keyring
but the built-in keys and secondary keyring are not used.

Add support for the built-in keys and secondary keyring as x86 does.

Fixes: e23a8020ce4e ("s390/kexec_file: Signature verification prototype")
Cc: stable@vger.kernel.org
Cc: Philipp Rudo <prudo@linux.ibm.com>
Cc: kexec@lists.infradead.org
Cc: keyrings@vger.kernel.org
Cc: linux-security-module@vger.kernel.org
Signed-off-by: Michal Suchanek <msuchanek@suse.de>
Reviewed-by: "Lee, Chun-Yi" <jlee@suse.com>
Acked-by: Baoquan He <bhe@redhat.com>
Signed-off-by: Coiby Xu <coxu@redhat.com>
Acked-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dm writecache: set a default MAX_WRITEBACK_JOBS

[ Upstream commit ca7dc242e358e46d963b32f9d9dd829785a9e957 ]

dm-writecache has the capability to limit the number of writeback jobs
in progress. However, this feature was off by default. As such there
were some out-of-memory crashes observed when lowering the low
watermark while the cache is full.

This commit enables writeback limit by default. It is set to 256MiB or
1/16 of total system memory, whichever is smaller.

Cc: stable@vger.kernel.org
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tty: 8250: Add support for Brainboxes PX cards.

[ Upstream commit ef5a03a26c87a760bc3d86b5af7b773e82f8b1b7 ]

Add support for some of the Brainboxes PCIe (PX) range of
serial cards, including the PX-101, PX-235/PX-246,
PX-203/PX-257, PX-260/PX-701, PX-310, PX-313,
PX-320/PX-324/PX-376/PX-387, PX-335/PX-346, PX-368, PX-420,
PX-803 and PX-846.

Signed-off-by: Cameron Williams <cang1@live.co.uk>
Cc: stable <stable@kernel.org>
Link: https://lore.kernel.org/r/AM5PR0202MB2564669252BDC59BF55A6E87C4879@AM5PR0202MB2564.eurprd02.prod.outlook.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

serial: 8250: Add proper clock handling for OxSemi PCIe devices

[ Upstream commit 366f6c955d4d1a5125ffcd6875ead26a3c7a2a1c ]

Oxford Semiconductor PCIe (Tornado) 950 serial port devices are driven
by a fixed 62.5MHz clock input derived from the 100MHz PCI Express clock.

We currently drive the device using its default oversampling rate of 16
and the clock prescaler disabled, consequently yielding the baud base of
3906250.  This base is inadequate for some of the high-speed baud rates
such as 460800bps, for which the closest rate possible can be obtained
by dividing the baud base by 8, yielding the baud rate of 488281.25bps,
which is off by 5.9638%.  This is enough for data communication to break
with the remote end talking actual 460800bps, where missed stop bits
have been observed.

We can do better however, by taking advantage of a reduced oversampling
rate, which can be set to any integer value from 4 to 16 inclusive by
programming the TCR register, and by using the clock prescaler, which
can be set to any value from 1 to 63.875 in increments of 0.125 in the
CPR/CPR2 register pair.  The prescaler has to be explicitly enabled
though by setting bit 7 in the MCR or otherwise it is bypassed (in the
enhanced mode that we enable) as if the value of 1 was used.

Make use of these features then as follows:

- Set the baud base to 15625000, reflecting the minimum oversampling
  rate of 4 with the clock prescaler and divisor both set to 1.

- Override the `set_mctrl' and set the MCR shadow there so as to have
  MCR[7] always set and have the 8250 core propagate these settings.

- Override the `get_divisor' handler and determine a good combination of
  parameters by using a lookup table with predetermined value pairs of
  the oversampling rate and the clock prescaler and finding a pair that
  divides the input clock such that the quotient, when rounded to the
  nearest integer, deviates the least from the exact result.  Calculate
  the clock divisor accordingly.

  Scale the resulting oversampling rate (only by powers of two) if
  possible so as to maximise it, reducing the divisor accordingly, and
  avoid a divisor overflow for very low baud rates by scaling the
  oversampling rate and/or the prescaler even if that causes some
  accuracy loss.

  Also handle the historic spd_cust feature so as to allow one to set
  all the three parameters manually to arbitrary values, by keeping the
  low 16 bits for the divisor and then putting TCR in bits 19:16 and
  CPR/CPR2 in bits 28:20, sanitising the bit pattern supplied such as
  to clamp CPR/CPR2 values between 0.000 and 0.875 inclusive to 33.875.
  This preserves compatibility with any existing setups, that is where
  requesting a custom divisor that only has any bits set among the low
  16 the oversampling rate of 16 and the clock prescaler of 33.875 will
  be used as with the original 8250.

  Finally abuse the `frac' argument to store the determined bit patterns
  for the TCR, CPR and CPR2 registers.

- Override the `set_divisor' handler so as to set the TCR, CPR and CPR2
  registers from the `frac' value supplied.  Set the divisor as usual.

With the baud base set to 15625000 and the unsigned 16-bit UART_DIV_MAX
limitation imposed by `serial8250_get_baud_rate' standard baud rates
below 300bps become unavailable in the regular way, e.g. the rate of
200bps requires the baud base to be divided by 78125 and that is beyond
the unsigned 16-bit range.  The historic spd_cust feature can still be
used to obtain such rates if so required.

See Documentation/tty/device_drivers/oxsemi-tornado.rst for more details.

Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2204181519450.9383@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

serial: 8250: Fold EndRun device support into OxSemi Tornado code

[ Upstream commit 1f32c65bad24b9787d3e52843de375430e3df822 ]

The EndRun PTP/1588 dual serial port device is based on the Oxford
Semiconductor OXPCIe952 UART device with the PCI vendor:device ID set
for EndRun Technologies and uses the same sequence to determine the
number of ports available.  Despite that we have duplicate code
specific to the EndRun device.

Remove redundant code then and factor out OxSemi Tornado device
detection.

Signed-off-by: Maciej W. Rozycki <macro@orcam.me.uk>
Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Link: https://lore.kernel.org/r/alpine.DEB.2.21.2204181516220.9383@angie.orcam.me.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

serial: 8250_pci: Replace dev_*() by pci_*() macros

[ Upstream commit 1177384179416c7136e1348f07609e0da1ae6b91 ]

PCI subsystem provides convenient shortcut macros for message printing.
Use those macros instead of dev_*().

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Jiri Slaby <jslaby@kernel.org>
Link: https://lore.kernel.org/r/20211022135147.70965-3-andriy.shevchenko@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

serial: 8250_pci: Refactor the loop in pci_ite887x_init()

[ Upstream commit 35b4f17231923e2f64521bdf7a2793ce2c3c74a6 ]

The loop can be refactored by using ARRAY_SIZE() instead of NULL terminator.
This reduces code base and makes it easier to read and understand.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Jiri Slaby <jslaby@kernel.org>
Link: https://lore.kernel.org/r/20211022135147.70965-1-andriy.shevchenko@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

PCI: qcom: Power on PHY before IPQ8074 DBI register accesses

[ Upstream commit a0e43bb9973b06ce5c666f0901e104e2037c1b34 ]

Currently the Gen2 port in IPQ8074 will cause the system to hang as it
accesses DBI registers in qcom_pcie_init_2_3_3(), and those are only
accesible after phy_power_on().

Move the DBI read/writes to a new qcom_pcie_post_init_2_3_3(), which is
executed after phy_power_on().

Link: https://lore.kernel.org/r/20220623155004.688090-1-robimarko@gmail.com
Fixes: a0fd361db8e5 ("PCI: dwc: Move "dbi", "dbi2", and "addr_space" resource setup into common code")
Signed-off-by: Robert Marko <robimarko@gmail.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Cc: stable@vger.kernel.org # v5.11+
Signed-off-by: Sasha Levin <sashal@kernel.org>

PCI/AER: Iterate over error counters instead of error strings

[ Upstream commit 5e6ae050955b566484f3cc6a66e3925eae87a0ed ]

Previously we iterated over AER stat *names*, e.g.,
aer_correctable_error_string[32], but the actual stat *counters* may not be
that large, e.g., pdev->aer_stats->dev_cor_errs[16], which means that we
printed junk in the sysfs stats files.

Iterate over the stat counter arrays instead of the names to avoid this
junk.

Also, added a build time check to make sure all
counters have entries in strings array.

Fixes: 0678e3109a3c ("PCI/AER: Simplify __aer_print_error()")
Link: https://lore.kernel.org/r/20220509181441.31884-1-mkhalfella@purestorage.com
Reported-by: Meeta Saggi <msaggi@purestorage.com>
Signed-off-by: Mohamed Khalfella <mkhalfella@purestorage.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Reviewed-by: Meeta Saggi <msaggi@purestorage.com>
Reviewed-by: Eric Badger <ebadger@purestorage.com>
Cc: stable@vger.kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

iommu/vt-d: avoid invalid memory access via node_online(NUMA_NO_NODE)

[ Upstream commit b0b0b77ea611e3088e9523e60860f4f41b62b235 ]

KASAN reports:

[ 4.668325][ T0] BUG: KASAN: wild-memory-access in dmar_parse_one_rhsa (arch/x86/include/asm/bitops.h:214 arch/x86/include/asm/bitops.h:226 include/asm-generic/bitops/instrumented-non-atomic.h:142 include/linux/nodemask.h:415 drivers/iommu/intel/dmar.c:497)
[    4.676149][    T0] Read of size 8 at addr 1fffffff85115558 by task swapper/0/0
[    4.683454][    T0]
[    4.685638][    T0] CPU: 0 PID: 0 Comm: swapper/0 Not tainted 5.19.0-rc3-00004-g0e862838f290 #1
[    4.694331][    T0] Hardware name: Supermicro SYS-5018D-FN4T/X10SDV-8C-TLN4F, BIOS 1.1 03/02/2016
[    4.703196][    T0] Call Trace:
[    4.706334][    T0]  <TASK>
[ 4.709133][ T0] ? dmar_parse_one_rhsa (arch/x86/include/asm/bitops.h:214 arch/x86/include/asm/bitops.h:226 include/asm-generic/bitops/instrumented-non-atomic.h:142 include/linux/nodemask.h:415 drivers/iommu/intel/dmar.c:497)

after converting the type of the first argument (@nr, bit number)
of arch_test_bit() from `long` to `unsigned long`[0].

Under certain conditions (for example, when ACPI NUMA is disabled
via command line), pxm_to_node() can return %NUMA_NO_NODE (-1).
It is valid 'magic' number of NUMA node, but not valid bit number
to use in bitops.
node_online() eventually descends to test_bit() without checking
for the input, assuming it's on caller side (which might be good
for perf-critical tasks). There, -1 becomes %ULONG_MAX which leads
to an insane array index when calculating bit position in memory.

For now, add an explicit check for @node being not %NUMA_NO_NODE
before calling test_bit(). The actual logics didn't change here
at all.

[0] https://github.com/norov/linux/commit/0e862838f290147ea9c16db852d8d494b552d38d

Fixes: ee34b32d8c29 ("dmar: support for parsing Remapping Hardware Static Affinity structure")
Cc: stable@vger.kernel.org # 2.6.33+
Reported-by: kernel test robot <oliver.sang@intel.com>
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Reviewed-by: Lu Baolu <baolu.lu@linux.intel.com>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

KVM: x86: Signal #GP, not -EPERM, on bad WRMSR(MCi_CTL/STATUS)

[ Upstream commit 2368048bf5c2ec4b604ac3431564071e89a0bc71 ]

Return '1', not '-1', when handling an illegal WRMSR to a MCi_CTL or
MCi_STATUS MSR.  The behavior of "all zeros' or "all ones" for CTL MSRs
is architectural, as is the "only zeros" behavior for STATUS MSRs.  I.e.
the intent is to inject a #GP, not exit to userspace due to an unhandled
emulation case.  Returning '-1' gets interpreted as -EPERM up the stack
and effecitvely kills the guest.

Fixes: 890ca9aefa78 ("KVM: Add MCE support")
Fixes: 9ffd986c6e4e ("KVM: X86: #GP when guest attempts to write MCi_STATUS register w/o 0")
Cc: stable@vger.kernel.org
Signed-off-by: Sean Christopherson <seanjc@google.com>
Reviewed-by: Jim Mattson <jmattson@google.com>
Link: https://lore.kernel.org/r/20220512222716.4112548-2-seanjc@google.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

KVM: set_msr_mce: Permit guests to ignore single-bit ECC errors

[ Upstream commit 0471a7bd1bca2a47a5f378f2222c5cf39ce94152 ]

Certain guest operating systems (e.g., UNIXWARE) clear bit 0 of
MC1_CTL to ignore single-bit ECC data errors.  Single-bit ECC data
errors are always correctable and thus are safe to ignore because they
are informational in nature rather than signaling a loss of data
integrity.

Prior to this patch, these guests would crash upon writing MC1_CTL,
with resultant error messages like the following:

error: kvm run failed Operation not permitted
EAX=fffffffe EBX=fffffffe ECX=00000404 EDX=ffffffff
ESI=ffffffff EDI=00000001 EBP=fffdaba4 ESP=fffdab20
EIP=c01333a5 EFL=00000246 [---Z-P-] CPL=0 II=0 A20=1 SMM=0 HLT=0
ES =0108 00000000 ffffffff 00c09300 DPL=0 DS   [-WA]
CS =0100 00000000 ffffffff 00c09b00 DPL=0 CS32 [-RA]
SS =0108 00000000 ffffffff 00c09300 DPL=0 DS   [-WA]
DS =0108 00000000 ffffffff 00c09300 DPL=0 DS   [-WA]
FS =0000 00000000 ffffffff 00c00000
GS =0000 00000000 ffffffff 00c00000
LDT=0118 c1026390 00000047 00008200 DPL=0 LDT
TR =0110 ffff5af0 00000067 00008b00 DPL=0 TSS32-busy
GDT=     ffff5020 000002cf
IDT=     ffff52f0 000007ff
CR0=8001003b CR2=00000000 CR3=0100a000 CR4=00000230
DR0=00000000 DR1=00000000 DR2=00000000 DR3=00000000
DR6=ffff0ff0 DR7=00000400
EFER=0000000000000000
Code=08 89 01 89 51 04 c3 8b 4c 24 08 8b 01 8b 51 04 8b 4c 24 04 <0f>
30 c3 f7 05 a4 6d ff ff 10 00 00 00 74 03 0f 31 c3 33 c0 33 d2 c3 8d
74 26 00 0f 31 c3

Signed-off-by: Lev Kujawski <lkujaw@member.fsf.org>
Message-Id: <20220521081511.187388-1-lkujaw@member.fsf.org>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

intel_th: pci: Add Raptor Lake-S CPU support

[ Upstream commit ff46a601afc5a66a81c3945b83d0a2caeb88e8bc ]

Add support for the Trace Hub in Raptor Lake-S CPU.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: stable <stable@kernel.org>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Link: https://lore.kernel.org/r/20220705082637.59979-7-alexander.shishkin@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

intel_th: pci: Add Raptor Lake-S PCH support

[ Upstream commit 23e2de5826e2fc4dd43e08bab3a2ea1a5338b063 ]

Add support for the Trace Hub in Raptor Lake-S PCH.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: stable <stable@kernel.org>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Link: https://lore.kernel.org/r/20220705082637.59979-6-alexander.shishkin@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

intel_th: pci: Add Meteor Lake-P support

[ Upstream commit 802a9a0b1d91274ef10d9fe429b4cc1e8c200aef ]

Add support for the Trace Hub in Meteor Lake-P.

Reviewed-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: stable <stable@kernel.org>
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Link: https://lore.kernel.org/r/20220705082637.59979-5-alexander.shishkin@linux.intel.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: lpfc: Remove extra atomic_inc on cmd_pending in queuecommand after VMID

[ Upstream commit 0948a9c5386095baae4012190a6b65aba684a907 ]

VMID introduced an extra increment of cmd_pending, causing double-counting
of the I/O. The normal increment ios performed in lpfc_get_scsi_buf.

Link: https://lore.kernel.org/r/20220701211425.2708-5-jsmart2021@gmail.com
Fixes: 33c79741deaf ("scsi: lpfc: vmid: Introduce VMID in I/O path")
Cc: <stable@vger.kernel.org> # v5.14+
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: lpfc: SLI path split: Refactor SCSI paths

[ Upstream commit 3512ac0942938d6977e7999ee69765d948d2faf1 ]

This patch refactors the SCSI paths to use SLI-4 as the primary interface.

 - Conversion away from using SLI-3 iocb structures to set/access fields in
   common routines. Use the new generic get/set routines that were added.
   This move changes code from indirect structure references to using local
   variables with the generic routines.

 - Refactor routines when setting non-generic fields, to have both SLI3 and
   SLI4 specific sections. This replaces the set-as-SLI3 then translate to
   SLI4 behavior of the past.

Link: https://lore.kernel.org/r/20220225022308.16486-14-jsmart2021@gmail.com
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: lpfc: SLI path split: Refactor fast and slow paths to native SLI4

[ Upstream commit 1b64aa9eae28ac598a03ed3d62a63ac5e5b295fc ]

Convert the SLI4 fast and slow paths to use native SLI4 wqe constructs
instead of iocb SLI3-isms.

Includes the following:

 - Create simple get_xxx and set_xxx routines to wrapper access to common
   elements in both SLI3 and SLI4 commands - allowing calling routines to
   avoid sli-rev-specific structures to access the elements.

 - using the wqe in the job structure as the primary element

 - use defines from SLI-4, not SLI-3

 - Removal of iocb to wqe conversion from fast and slow path

 - Add below routines to handle fast path
lpfc_prep_embed_io - prepares the wqe for fast path
lpfc_wqe_bpl2sgl   - manages bpl to sgl conversion
lpfc_sli_wqe2iocb  - converts a WQE to IOCB for SLI-3 path

 - Add lpfc_sli3_iocb2wcqecmpl in completion path to convert an SLI-3
   iocb completion to wcqe completion

 - Refactor some of the code that works on both revs for clarity

Link: https://lore.kernel.org/r/20220225022308.16486-3-jsmart2021@gmail.com
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: lpfc: SLI path split: Refactor lpfc_iocbq

[ Upstream commit a680a9298e7b4ff344aca3456177356b276e5038 ]

Currently, SLI3 and SLI4 data paths use the same lpfc_iocbq structure.
This is a "common" structure but many of the components refer to sli-rev
specific entities which can lead the developer astray as to what they
actually mean, should be set to, or when they should be used.

This first patch prepares the lpfc_iocbq structure so that elements common
to both SLI3 and SLI4 data paths are more appropriately named, making it
clear they apply generically.

Fieldnames based on 'iocb' (sli3) or 'wqe' (sli4) which are actually
generic to the paths are renamed to 'cmd':

 - iocb_flag is renamed to cmd_flag

 - lpfc_vmid_iocb_tag is renamed to lpfc_vmid_tag

 - fabric_iocb_cmpl is renamed to fabric_cmd_cmpl

 - wait_iocb_cmpl is renamed to wait_cmd_cmpl

 - iocb_cmpl and wqe_cmpl are combined and renamed to cmd_cmpl

 - rsvd2 member is renamed to num_bdes due to pre-existing usage

The structure name itself will retain the iocb reference as changing to a
more relevant "job" or "cmd" title induces many hundreds of line changes
for only a name change.

lpfc_post_buffer is also renamed to lpfc_sli3_post_buffer to indicate use
in the SLI3 path only.

Link: https://lore.kernel.org/r/20220225022308.16486-2-jsmart2021@gmail.com
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

scsi: lpfc: Fix EEH support for NVMe I/O

[ Upstream commit 25ac2c970be32993f1dff607f8354f3c053d42bc ]

Injecting errors on the PCI slot while the driver is handling NVMe I/O will
cause crashes and hangs.

There are several rather difficult scenarios occurring. The main issue is
that the adapter can report a PCI error before or simultaneously to the PCI
subsystem reporting the error. Both paths have different entry points and
currently there is no interlock between them. Thus multiple teardown paths
are competing and all heck breaks loose.

Complicating things is the NVMs path. To a large degree, I/O was able to be
shutdown for a full FC port on the SCSI stack. But on NVMe, there isn't a
similar call. At best, it works on a per-controller basis, but even at the
controller level, it's a controller "reset" call. All of which means I/O is
still flowing on different CPUs with reset paths expecting hw access
(mailbox commands) to execute properly.

The following modifications are made:

 - A new flag is set in PCI error entrypoints so the driver can track being
   called by that path.

 - An interlock is added in the SLI hw error path and the PCI error path
   such that only one of the paths proceeds with the teardown logic.

 - RPI cleanup is patched such that RPIs are marked unregistered w/o mbx
   cmds in cases of hw error.

 - If entering the SLI port re-init calls, a case where SLI error teardown
   was quick and beat the PCI calls now reporting error, check whether the
   SLI port is still live on the PCI bus.

 - In the PCI reset code to bring the adapter back, recheck the IRQ
   settings. Different checks for SLI3 vs SLI4.

 - In I/O completions, that may be called as part of the cleanup or
   underway just before the hw error, check the state of the adapter.  If
   in error, shortcut handling that would expect further adapter
   completions as the hw error won't be sending them.

 - In routines waiting on I/O completions, which may have been in progress
   prior to the hw error, detect the device is being torn down and abort
   from their waits and just give up. This points to a larger issue in the
   driver on ref-counting for data structures, as it doesn't have
   ref-counting on q and port structures. We'll do this fix for now as it
   would be a major rework to be done differently.

 - Fix the NVMe cleanup to simulate NVMe I/O completions if I/O is being
   failed back due to hw error.

 - In I/O buf allocation, done at the start of new I/Os, check hw state and
   fail if hw error.

Link: https://lore.kernel.org/r/20210910233159.115896-10-jsmart2021@gmail.com
Co-developed-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: Justin Tee <justin.tee@broadcom.com>
Signed-off-by: James Smart <jsmart2021@gmail.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

firmware: arm_scpi: Ensure scpi_info is not assigned if the probe fails

[ Upstream commit 689640efc0a2c4e07e6f88affe6d42cd40cc3f85 ]

When scpi probe fails, at any point, we need to ensure that the scpi_info
is not set and will remain NULL until the probe succeeds. If it is not
taken care, then it could result use-after-free as the value is exported
via get_scpi_ops() and could refer to a memory allocated via devm_kzalloc()
but freed when the probe fails.

Link: https://lore.kernel.org/r/20220701160310.148344-1-sudeep.holla@arm.com
Cc: stable@vger.kernel.org # 4.19+
Reported-by: huhai <huhai@kylinos.cn>
Reviewed-by: Jackie Liu <liuyun01@kylinos.cn>
Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

usbnet: smsc95xx: Fix deadlock on runtime resume

[ Upstream commit 7b960c967f2aa01ab8f45c5a0bd78e754cffdeee ]

Commit 05b35e7eb9a1 ("smsc95xx: add phylib support") amended
smsc95xx_resume() to call phy_init_hw().  That function waits for the
device to runtime resume even though it is placed in the runtime resume
path, causing a deadlock.

The problem is that phy_init_hw() calls down to smsc95xx_mdiobus_read(),
which never uses the _nopm variant of usbnet_read_cmd().

Commit b4df480f68ae ("usbnet: smsc95xx: add reset_resume function with
reset operation") causes a similar deadlock on resume if the device was
already runtime suspended when entering system sleep:

That's because the commit introduced smsc95xx_reset_resume(), which
calls down to smsc95xx_reset(), which neglects to use _nopm accessors.

Fix by auto-detecting whether a device access is performed by the
suspend/resume task_struct and use the _nopm variant if so.  This works
because the PM core guarantees that suspend/resume callbacks are run in
task context.

Stacktrace for posterity:

  INFO: task kworker/2:1:49 blocked for more than 122 seconds.
  Workqueue: usb_hub_wq hub_event
  schedule
  rpm_resume
  __pm_runtime_resume
  usb_autopm_get_interface
  usbnet_read_cmd
  __smsc95xx_read_reg
  __smsc95xx_phy_wait_not_busy
  __smsc95xx_mdio_read
  smsc95xx_mdiobus_read
  __mdiobus_read
  mdiobus_read
  smsc_phy_reset
  phy_init_hw
  smsc95xx_resume
  usb_resume_interface
  usb_resume_both
  usb_runtime_resume
  __rpm_callback
  rpm_callback
  rpm_resume
  __pm_runtime_resume
  usb_autoresume_device
  hub_event
  process_one_work

Fixes: b4df480f68ae ("usbnet: smsc95xx: add reset_resume function with reset operation")
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Cc: stable@vger.kernel.org # v3.16+
Cc: Andre Edich <andre.edich@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

usbnet: smsc95xx: Forward PHY interrupts to PHY driver to avoid polling

[ Upstream commit 1ce8b37241ed291af56f7a49bbdbf20c08728e88 ]

Link status of SMSC LAN95xx chips is polled once per second, even though
they're capable of signaling PHY interrupts through the MAC layer.

Forward those interrupts to the PHY driver to avoid polling.  Benefits
are reduced bus traffic, reduced CPU overhead and quicker interface
bringup.

Polling was introduced in 2016 by commit d69d16949346 ("usbnet:
smsc95xx: fix link detection for disabled autonegotiation").
Back then, the LAN95xx driver neglected to enable the ENERGYON interrupt,
hence couldn't detect link-up events when auto-negotiation was disabled.
The proper solution would have been to enable the ENERGYON interrupt
instead of polling.

Since then, PHY handling was moved from the LAN95xx driver to the SMSC
PHY driver with commit 05b35e7eb9a1 ("smsc95xx: add phylib support").
That PHY driver is capable of link detection with auto-negotiation
disabled because it enables the ENERGYON interrupt.

Note that signaling interrupts through the MAC layer not only works with
the integrated PHY, but also with an external PHY, provided its
interrupt pin is attached to LAN95xx's nPHY_INT pin.

In the unlikely event that the interrupt pin of an external PHY is
attached to a GPIO of the SoC (or not connected at all), the driver can
be amended to retrieve the irq from the PHY's of_node.

To forward PHY interrupts to phylib, it is not sufficient to call
phy_mac_interrupt().  Instead, the PHY's interrupt handler needs to run
so that PHY interrupts are cleared.  That's because according to page
119 of the LAN950x datasheet, "The source of this interrupt is a level.
The interrupt persists until it is cleared in the PHY."

https://www.microchip.com/content/dam/mchp/documents/UNG/ProductDocuments/DataSheets/LAN950x-Data-Sheet-DS00001875D.pdf

Therefore, create an IRQ domain with a single IRQ for the PHY.  In the
future, the IRQ domain may be extended to support the 11 GPIOs on the
LAN95xx.

Normally the PHY interrupt should be masked until the PHY driver has
cleared it.  However masking requires a (sleeping) USB transaction and
interrupts are received in (non-sleepable) softirq context.  I decided
not to mask the interrupt at all (by using the dummy_irq_chip's noop
->irq_mask() callback):  The USB interrupt endpoint is polled in 1 msec
intervals and normally that's sufficient to wake the PHY driver's IRQ
thread and have it clear the interrupt.  If it does take longer, worst
thing that can happen is the IRQ thread is woken again.  No big deal.

Because PHY interrupts are now perpetually enabled, there's no need to
selectively enable them on suspend.  So remove all invocations of
smsc95xx_enable_phy_wakeup_interrupts().

In smsc95xx_resume(), move the call of phy_init_hw() before
usbnet_resume() (which restarts the status URB) to ensure that the PHY
is fully initialized when an interrupt is handled.

Tested-by: Oleksij Rempel <o.rempel@pengutronix.de> # LAN9514/9512/9500
Tested-by: Ferry Toth <fntoth@gmail.com> # LAN9514
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch> # from a PHY perspective
Cc: Andre Edich <andre.edich@microchip.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

usbnet: smsc95xx: Avoid link settings race on interrupt reception

[ Upstream commit 8960f878e39fadc03d74292a6731f1e914cf2019 ]

When a PHY interrupt is signaled, the SMSC LAN95xx driver updates the
MAC full duplex mode and PHY flow control registers based on cached data
in struct phy_device:

  smsc95xx_status()                 # raises EVENT_LINK_RESET
    usbnet_deferred_kevent()
      smsc95xx_link_reset()         # uses cached data in phydev

Simultaneously, phylib polls link status once per second and updates
that cached data:

  phy_state_machine()
    phy_check_link_status()
      phy_read_status()
        lan87xx_read_status()
          genphy_read_status()      # updates cached data in phydev

If smsc95xx_link_reset() wins the race against genphy_read_status(),
the registers may be updated based on stale data.

E.g. if the link was previously down, phydev->duplex is set to
DUPLEX_UNKNOWN and that's what smsc95xx_link_reset() will use, even
though genphy_read_status() may update it to DUPLEX_FULL afterwards.

PHY interrupts are currently only enabled on suspend to trigger wakeup,
so the impact of the race is limited, but we're about to enable them
perpetually.

Avoid the race by delaying execution of smsc95xx_link_reset() until
phy_state_machine() has done its job and calls back via
smsc95xx_handle_link_change().

Signaling EVENT_LINK_RESET on wakeup is not necessary because phylib
picks up link status changes through polling.  So drop the declaration
of a ->link_reset() callback.

Note that the semicolon on a line by itself added in smsc95xx_status()
is a placeholder for a function call which will be added in a subsequent
commit.  That function call will actually handle the INT_ENP_PHY_INT_
interrupt.

Tested-by: Oleksij Rempel <o.rempel@pengutronix.de> # LAN9514/9512/9500
Tested-by: Ferry Toth <fntoth@gmail.com> # LAN9514
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

usbnet: smsc95xx: Don't clear read-only PHY interrupt

[ Upstream commit 3108871f19221372b251f7da1ac38736928b5b3a ]

Upon receiving data from the Interrupt Endpoint, the SMSC LAN95xx driver
attempts to clear the signaled interrupts by writing "all ones" to the
Interrupt Status Register.

However the driver only ever enables a single type of interrupt, namely
the PHY Interrupt.  And according to page 119 of the LAN950x datasheet,
its bit in the Interrupt Status Register is read-only.  There's no other
way to clear it than in a separate PHY register:

https://www.microchip.com/content/dam/mchp/documents/UNG/ProductDocuments/DataSheets/LAN950x-Data-Sheet-DS00001875D.pdf

Consequently, writing "all ones" to the Interrupt Status Register is
pointless and can be dropped.

Tested-by: Oleksij Rempel <o.rempel@pengutronix.de> # LAN9514/9512/9500
Tested-by: Ferry Toth <fntoth@gmail.com> # LAN9514
Signed-off-by: Lukas Wunner <lukas@wunner.de>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/vc4: drv: Adopt the dma configuration from the HVS or V3D component

[ Upstream commit da8e393e23efb60eba8959856c7df88f9859f6eb ]

vc4_drv isn't necessarily under the /soc node in DT as it is a
virtual device, but it is the one that does the allocations.
The DMA addresses are consumed by primarily the HVS or V3D, and
those require VideoCore cache alias address mapping, and so will be
under /soc.

During probe find the a suitable device node for HVS or V3D,
and adopt the DMA configuration of that node.

Cc: <stable@vger.kernel.org>
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.com>
Link: https://lore.kernel.org/r/20220613144800.326124-2-maxime@cerno.tech
Signed-off-by: Maxime Ripard <maxime@cerno.tech>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/dp/mst: Read the extended DPCD capabilities during system resume

[ Upstream commit 7a710a8bc909313951eb9252d8419924c771d7c2 ]

The WD22TB4 Thunderbolt dock at least will revert its DP_MAX_LINK_RATE
from HBR3 to HBR2 after system suspend/resume if the DP_DP13_DPCD_REV
registers are not read subsequently also as required.

Fix this by reading DP_DP13_DPCD_REV registers as well, matching what is
done during connector detection. While at it also fix up the same call
in drm_dp_mst_dump_topology().

Cc: Lyude Paul <lyude@redhat.com>
Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/5292
Signed-off-by: Imre Deak <imre.deak@intel.com>
Reviewed-by: Jani Nikula <jani.nikula@intel.com>
Cc: <stable@vger.kernel.org> # v5.14+
Reviewed-by: Lyude Paul <lyude@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20220614094537.885472-1-imre.deak@intel.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

crypto: blake2s - remove shash module

[ Upstream commit 2d16803c562ecc644803d42ba98a8e0aef9c014e ]

BLAKE2s has no currently known use as an shash. Just remove all of this
unnecessary plumbing. Removing this shash was something we talked about
back when we were making BLAKE2s a built-in, but I simply never got
around to doing it. So this completes that project.

Importantly, this fixs a bug in which the lib code depends on
crypto_simd_disabled_for_test, causing linker errors.

Also add more alignment tests to the selftests and compare SIMD and
non-SIMD compression functions, to make up for what we lose from
testmgr.c.

Reported-by: gaochao <gaochao49@huawei.com>
Cc: Eric Biggers <ebiggers@kernel.org>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: stable@vger.kernel.org
Fixes: 6048fdcc5f26 ("lib/crypto: blake2s: include as built-in")
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/mediatek: Keep dsi as LP00 before dcs cmds transfer

[ Upstream commit 39e8d062b03c3dc257d880d82bd55cdd9e185a3b ]

To comply with the panel sequence, hold the mipi signal to LP00 before
the dcs cmds transmission, and pull the mipi signal high from LP00 to
LP11 until the start of the dcs cmds transmission.

The normal panel timing is :
(1) pp1800 DC pull up
(2) avdd & avee AC pull high
(3) lcm_reset pull high -> pull low -> pull high
(4) Pull MIPI signal high (LP11) -> initial code -> send video data
    (HS mode)

The power-off sequence is reversed.
If dsi is not in cmd mode, then dsi will pull the mipi signal high in
the mtk_output_dsi_enable function. The delay in lane_ready func is
the reaction time of dsi_rx after pulling up the mipi signal.

Fixes: 2dd8075d2185 ("drm/mediatek: mtk_dsi: Use the drm_panel_bridge API")

Link: https://patchwork.kernel.org/project/linux-mediatek/patch/1653012007-11854-4-git-send-email-xinlei.lee@mediatek.com/
Cc: <stable@vger.kernel.org> # 5.10.x: 7f6335c6a258: drm/mediatek: Modify dsi funcs to atomic operations
Cc: <stable@vger.kernel.org> # 5.10.x: cde7e2e35c28: drm/mediatek: Separate poweron/poweroff from enable/disable and define new funcs
Cc: <stable@vger.kernel.org> # 5.10.x
Signed-off-by: Jitao Shi <jitao.shi@mediatek.com>
Signed-off-by: Xinlei Lee <xinlei.lee@mediatek.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Reviewed-by: Rex-BC Chen <rex-bc.chen@mediatek.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/mediatek: Allow commands to be sent during video mode

[ Upstream commit 81cc7e51c4f1686b71e30046437056ece6b2cb4d ]

Mipi dsi panel drivers can use mipi_dsi_dcs_{set,get}_display_brightness()
to request backlight changes.

This can be done during panel initialization (dsi is in command mode)
or afterwards (dsi is in Video Mode).

When the DSI is in Video Mode, all commands are rejected.

Detect current DSI mode in mtk_dsi_host_transfer() and switch modes
temporarily to allow commands to be sent.

Signed-off-by: Julien STEPHAN <jstephan@baylibre.com>
Signed-off-by: Mattijs Korpershoek <mkorpershoek@baylibre.com>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Chun-Kuang Hu <chunkuang.hu@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spmi: trace: fix stack-out-of-bound access in SPMI tracing functions

commit 2af28b241eea816e6f7668d1954f15894b45d7e3 upstream.

trace_spmi_write_begin() and trace_spmi_read_end() both call
memcpy() with a length of "len + 1".  This leads to one extra
byte being read beyond the end of the specified buffer.  Fix
this out-of-bound memory access by using a length of "len"
instead.

Here is a KASAN log showing the issue:

BUG: KASAN: stack-out-of-bounds in trace_event_raw_event_spmi_read_end+0x1d0/0x234
Read of size 2 at addr ffffffc0265b7540 by task thermal@2.0-ser/1314
...
Call trace:
 dump_backtrace+0x0/0x3e8
 show_stack+0x2c/0x3c
 dump_stack_lvl+0xdc/0x11c
 print_address_description+0x74/0x384
 kasan_report+0x188/0x268
 kasan_check_range+0x270/0x2b0
 memcpy+0x90/0xe8
 trace_event_raw_event_spmi_read_end+0x1d0/0x234
 spmi_read_cmd+0x294/0x3ac
 spmi_ext_register_readl+0x84/0x9c
 regmap_spmi_ext_read+0x144/0x1b0 [regmap_spmi]
 _regmap_raw_read+0x40c/0x754
 regmap_raw_read+0x3a0/0x514
 regmap_bulk_read+0x418/0x494
 adc5_gen3_poll_wait_hs+0xe8/0x1e0 [qcom_spmi_adc5_gen3]
 ...
 __arm64_sys_read+0x4c/0x60
 invoke_syscall+0x80/0x218
 el0_svc_common+0xec/0x1c8
 ...

addr ffffffc0265b7540 is located in stack of task thermal@2.0-ser/1314 at offset 32 in frame:
 adc5_gen3_poll_wait_hs+0x0/0x1e0 [qcom_spmi_adc5_gen3]

this frame has 1 object:
 [32, 33) 'status'

Memory state around the buggy address:
 ffffffc0265b7400: 00 00 00 00 00 00 00 00 00 00 00 00 f1 f1 f1 f1
 ffffffc0265b7480: 04 f3 f3 f3 00 00 00 00 00 00 00 00 00 00 00 00
>ffffffc0265b7500: 00 00 00 00 f1 f1 f1 f1 01 f3 f3 f3 00 00 00 00
                                           ^
 ffffffc0265b7580: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 ffffffc0265b7600: f1 f1 f1 f1 01 f2 07 f2 f2 f2 01 f3 00 00 00 00
==================================================================

Fixes: a9fce374815d ("spmi: add command tracepoints for SPMI")
Cc: stable@vger.kernel.org
Reviewed-by: Stephen Boyd <sboyd@kernel.org>
Acked-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: David Collins <quic_collinsd@quicinc.com>
Link: https://lore.kernel.org/r/20220627235512.2272783-1-quic_collinsd@quicinc.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

__follow_mount_rcu(): verify that mount_lock remains unchanged

commit 20aac6c60981f5bfacd66661d090d907bf1482f0 upstream.

Validate mount_lock seqcount as soon as we cross into mount in RCU
mode.  Sure, ->mnt_root is pinned and will remain so until we
do rcu_read_unlock() anyway, and we will eventually fail to unlazy if
the mount_lock had been touched, but we might run into a hard error
(e.g. -ENOENT) before trying to unlazy.  And it's possible to end
up with RCU pathwalk racing with rename() and umount() in a way
that would fail with -ENOENT while non-RCU pathwalk would've
succeeded with any timings.

Once upon a time we hadn't needed that, but analysis had been subtle,
brittle and went out of window as soon as RENAME_EXCHANGE had been
added.

It's narrow, hard to hit and won't get you anything other than
stray -ENOENT that could be arranged in much easier way with the
same priveleges, but it's a bug all the same.

Cc: stable@kernel.org
X-sky-is-falling: unlikely
Fixes: da1ce0670c14 "vfs: add cross-rename"
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Input: gscps2 - check return value of ioremap() in gscps2_probe()

commit e61b3125a4f036b3c6b87ffd656fc1ab00440ae9 upstream.

The function ioremap() in gscps2_probe() can fail, so
its return value should be checked.

Fixes: 4bdc0d676a643 ("remove ioremap_nocache and devm_ioremap_nocache")
Cc: <stable@vger.kernel.org> # v5.6+
Reported-by: Hacash Robot <hacashRobot@santino.com>
Signed-off-by: Xie Shaowen <studentxswpy@163.com>
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

posix-cpu-timers: Cleanup CPU timers before freeing them during exec

commit e362359ace6f87c201531872486ff295df306d13 upstream.

Commit 55e8c8eb2c7b ("posix-cpu-timers: Store a reference to a pid not a
task") started looking up tasks by PID when deleting a CPU timer.

When a non-leader thread calls execve, it will switch PIDs with the leader
process. Then, as it calls exit_itimers, posix_cpu_timer_del cannot find
the task because the timer still points out to the old PID.

That means that armed timers won't be disarmed, that is, they won't be
removed from the timerqueue_list. exit_itimers will still release their
memory, and when that list is later processed, it leads to a
use-after-free.

Clean up the timers from the de-threaded task before freeing them. This
prevents a reported use-after-free.

Fixes: 55e8c8eb2c7b ("posix-cpu-timers: Store a reference to a pid not a task")
Signed-off-by: Thadeu Lima de Souza Cascardo <cascardo@canonical.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Thomas Gleixner <tglx@linutronix.de>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20220809170751.164716-1-cascardo@canonical.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

SMB3: fix lease break timeout when multiple deferred close handles for the same file.

commit 9e31678fb403eae0f4fe37c6374be098835c73cd upstream.

Solution is to send lease break ack immediately even in case of
deferred close handles to avoid lease break request timing out
and let deferred closed handle gets closed as scheduled.
Later patches could optimize cases where we then close some
of these handles sooner for the cases where lease break is to 'none'

Cc: stable@kernel.org
Signed-off-by: Bharath SM <bharathsm@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/olpc: fix 'logical not is only applied to the left hand side'

commit 3a2ba42cbd0b669ce3837ba400905f93dd06c79f upstream.

The bitops compile-time optimization series revealed one more
problem in olpc-xo1-sci.c:send_ebook_state(), resulted in GCC
warnings:

arch/x86/platform/olpc/olpc-xo1-sci.c: In function 'send_ebook_state':
arch/x86/platform/olpc/olpc-xo1-sci.c:83:63: warning: logical not is only applied to the left hand side of comparison [-Wlogical-not-parentheses]
   83 |         if (!!test_bit(SW_TABLET_MODE, ebook_switch_idev->sw) == state)
      |                                                               ^~
arch/x86/platform/olpc/olpc-xo1-sci.c:83:13: note: add parentheses around left hand side expression to silence this warning

Despite this code working as intended, this redundant double
negation of boolean value, together with comparing to `char`
with no explicit conversion to bool, makes compilers think
the author made some unintentional logical mistakes here.
Make it the other way around and negate the char instead
to silence the warnings.

Fixes: d2aa37411b8e ("x86/olpc/xo1/sci: Produce wakeup events for buttons and switches")
Cc: stable@vger.kernel.org # 3.5+
Reported-by: Guenter Roeck <linux@roeck-us.net>
Reported-by: kernel test robot <lkp@intel.com>
Reviewed-and-tested-by: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Alexander Lobakin <alexandr.lobakin@intel.com>
Signed-off-by: Yury Norov <yury.norov@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/kprobes: Update kcb status flag after singlestepping

commit dec8784c9088b131a1523f582c2194cfc8107dc0 upstream.

Fix kprobes to update kcb (kprobes control block) status flag to
KPROBE_HIT_SSDONE even if the kp->post_handler is not set.

This bug may cause a kernel panic if another INT3 user runs right
after kprobes because kprobe_int3_handler() misunderstands the
INT3 is kprobe's single stepping INT3.

Fixes: 6256e668b7af ("x86/kprobes: Use int3 instead of debug trap for single-step")
Reported-by: Daniel Müller <deso@posteo.net>
Signed-off-by: Masami Hiramatsu (Google) <mhiramat@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: Daniel Müller <deso@posteo.net>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/all/20220727210136.jjgc3lpqeq42yr3m@muellerd-fedora-PC2BDTX9
Link: https://lore.kernel.org/r/165942025658.342061.12452378391879093249.stgit@devnote2
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ftrace/x86: Add back ftrace_expected assignment

commit ac6c1b2ca77e722a1e5d651f12f437f2f237e658 upstream.

When a ftrace_bug happens (where ftrace fails to modify a location) it is
helpful to have what was at that location as well as what was expected to
be there.

But with the conversion to text_poke() the variable that assigns the
expected for debugging was dropped. Unfortunately, I noticed this when I
needed it. Add it back.

Link: https://lkml.kernel.org/r/20220726101851.069d2e70@gandalf.local.home
Cc: "x86@kernel.org" <x86@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Borislav Petkov <bp@alien8.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@vger.kernel.org
Fixes: 768ae4406a5c ("x86/ftrace: Use text_poke()")
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/bugs: Enable STIBP for IBPB mitigated RETBleed

commit e6cfcdda8cbe81eaf821c897369a65fec987b404 upstream.

AMD's "Technical Guidance for Mitigating Branch Type Confusion,
Rev. 1.0 2022-07-12" whitepaper, under section 6.1.2 "IBPB On
Privileged Mode Entry / SMT Safety" says:

  Similar to the Jmp2Ret mitigation, if the code on the sibling thread
  cannot be trusted, software should set STIBP to 1 or disable SMT to
  ensure SMT safety when using this mitigation.

So, like already being done for retbleed=unret, and now also for
retbleed=ibpb, force STIBP on machines that have it, and report its SMT
vulnerability status accordingly.

 [ bp: Remove the "we" and remove "[AMD]" applicability parameter which
   doesn't work here. ]

Fixes: 3ebc17006888 ("x86/bugs: Add retbleed=ibpb")
Signed-off-by: Kim Phillips <kim.phillips@amd.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: stable@vger.kernel.org # 5.10, 5.15, 5.19
Link: https://bugzilla.kernel.org/show_bug.cgi?id=206537
Link: https://lore.kernel.org/r/20220804192201.439596-1-kim.phillips@amd.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Fix losing FCP-2 targets during port perturbation tests

commit 58d1c124cd79ea686b512043c5bd515590b2ed95 upstream.

When a mix of FCP-2 (tape) and non-FCP-2 targets are present, FCP-2 target
state was incorrectly transitioned when both of the targets were gone. Fix
this by ignoring state transition for FCP-2 targets.

Link: https://lore.kernel.org/r/20220616053508.27186-7-njavali@marvell.com
Fixes: 44c57f205876 ("scsi: qla2xxx: Changes to support FCP2 Target")
Cc: stable@vger.kernel.org
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Fix losing target when it reappears during delete

commit 118b0c863c8f5629cc5271fc24d72d926e0715d9 upstream.

FC target disappeared during port perturbation tests due to a race that
tramples target state.  Fix the issue by adding state checks before
proceeding.

Link: https://lore.kernel.org/r/20220616053508.27186-8-njavali@marvell.com
Fixes: 44c57f205876 ("scsi: qla2xxx: Changes to support FCP2 Target")
Cc: stable@vger.kernel.org
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Fix losing FCP-2 targets on long port disable with I/Os

commit 2416ccd3815ba1613e10a6da0a24ef21acfe5633 upstream.

FCP-2 devices were not coming back online once they were lost, login
retries exhausted, and then came back up.  Fix this by accepting RSCN when
the device is not online.

Link: https://lore.kernel.org/r/20220616053508.27186-10-njavali@marvell.com
Fixes: 44c57f205876 ("scsi: qla2xxx: Changes to support FCP2 Target")
Cc: stable@vger.kernel.org
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Wind down adapter after PCIe error

commit d3117c83ba316b3200d9f2fe900f2b9a5525a25c upstream.

Put adapter into a wind down state if OS does not make any attempt to
recover the adapter after PCIe error.

Link: https://lore.kernel.org/r/20220616053508.27186-4-njavali@marvell.com
Cc: stable@vger.kernel.org
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Fix erroneous mailbox timeout after PCI error injection

commit f260694e6463b63ae550aad25ddefe94cb1904da upstream.

Clear wait for mailbox interrupt flag to prevent stale mailbox:

Feb 22 05:22:56 ltcden4-lp7 kernel: qla2xxx [0135:90:00.1]-500a:4: LOOP UP detected (16 Gbps).
Feb 22 05:22:59 ltcden4-lp7 kernel: qla2xxx [0135:90:00.1]-d04c:4: MBX Command timeout for cmd 69, ...

To fix the issue, driver needs to clear the MBX_INTR_WAIT flag on purging
the mailbox. When the stale mailbox completion does arrive, it will be
dropped.

Link: https://lore.kernel.org/r/20220616053508.27186-11-njavali@marvell.com
Fixes: b6faaaf796d7 ("scsi: qla2xxx: Serialize mailbox request")
Cc: Naresh Bannoth <nbannoth@in.ibm.com>
Cc: Kyle Mahlkuch <Kyle.Mahlkuch@ibm.com>
Cc: stable@vger.kernel.org
Reported-by: Naresh Bannoth <nbannoth@in.ibm.com>
Tested-by: Naresh Bannoth <nbannoth@in.ibm.com>
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Fix excessive I/O error messages by default

commit bff4873c709085e09d0ffae0c25b8e65256e3205 upstream.

Disable printing I/O error messages by default.  The messages will be
printed only when logging was enabled.

Link: https://lore.kernel.org/r/20220616053508.27186-2-njavali@marvell.com
Fixes: 8e2d81c6b5be ("scsi: qla2xxx: Fix excessive messages during device logout")
Cc: stable@vger.kernel.org
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Fix crash due to stale SRB access around I/O timeouts

commit c39587bc0abaf16593f7abcdf8aeec3c038c7d52 upstream.

Ensure SRB is returned during I/O timeout error escalation. If that is not
possible fail the escalation path.

Following crash stack was seen:

BUG: unable to handle kernel paging request at 0000002f56aa90f8
IP: qla_chk_edif_rx_sa_delete_pending+0x14/0x30 [qla2xxx]
Call Trace:
 ? qla2x00_status_entry+0x19f/0x1c50 [qla2xxx]
 ? qla2x00_start_sp+0x116/0x1170 [qla2xxx]
 ? dma_pool_alloc+0x1d6/0x210
 ? mempool_alloc+0x54/0x130
 ? qla24xx_process_response_queue+0x548/0x12b0 [qla2xxx]
 ? qla_do_work+0x2d/0x40 [qla2xxx]
 ? process_one_work+0x14c/0x390

Link: https://lore.kernel.org/r/20220616053508.27186-6-njavali@marvell.com
Fixes: d74595278f4a ("scsi: qla2xxx: Add multiple queue pair functionality.")
Cc: stable@vger.kernel.org
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Turn off multi-queue for 8G adapters

commit 5304673bdb1635e27555bd636fd5d6956f1cd552 upstream.

For 8G adapters, multi-queue was enabled accidentally. Make sure
multi-queue is not enabled.

Link: https://lore.kernel.org/r/20220616053508.27186-5-njavali@marvell.com
Cc: stable@vger.kernel.org
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Fix discovery issues in FC-AL topology

commit 47ccb113cead905bdc236571bf8ac6fed90321b3 upstream.

A direct attach tape device, when gets swapped with another, was not
discovered. Fix this by looking at loop map and reinitialize link if there
are devices present.

Link: https://lore.kernel.org/linux-scsi/baef87c3-5dad-3b47-44c1-6914bfc90108@cybernetics.com/
Link: https://lore.kernel.org/r/20220713052045.10683-8-njavali@marvell.com
Cc: stable@vger.kernel.org
Reported-by: Tony Battersby <tonyb@cybernetics.com>
Tested-by: Tony Battersby <tonyb@cybernetics.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Arun Easi <aeasi@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: qla2xxx: Fix imbalance vha->vref_count

commit 63fa7f2644b4b48e1913af33092c044bf48e9321 upstream.

vref_count took an extra decrement in the task management path.  Add an
extra ref count to compensate the imbalance.

Link: https://lore.kernel.org/r/20220713052045.10683-7-njavali@marvell.com
Cc: stable@vger.kernel.org
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Quinn Tran <qutran@marvell.com>
Signed-off-by: Nilesh Javali <njavali@marvell.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: zfcp: Fix missing auto port scan and thus missing target ports

commit 4da8c5f76825269f28d6a89fa752934a4bcb6dfa upstream.

Case (1):
  The only waiter on wka_port->completion_wq is zfcp_fc_wka_port_get()
  trying to open a WKA port. As such it should only be woken up by WKA port
  *open* responses, not by WKA port close responses.

Case (2):
  A close WKA port response coming in just after having sent a new open WKA
  port request and before blocking for the open response with wait_event()
  in zfcp_fc_wka_port_get() erroneously renders the wait_event a NOP
  because the close handler overwrites wka_port->status. Hence the
  wait_event condition is erroneously true and it does not enter blocking
  state.

With non-negligible probability, the following time space sequence happens
depending on timing without this fix:

user process        ERP thread zfcp work queue tasklet system work queue
============        ========== =============== ======= =================
$ echo 1 > online
zfcp_ccw_set_online
zfcp_ccw_activate
zfcp_erp_adapter_reopen
msleep scan backoff zfcp_erp_strategy
|                   ...
|                   zfcp_erp_action_cleanup
|                   ...
|                   queue delayed scan_work
|                   queue ns_up_work
|                              ns_up_work:
|                              zfcp_fc_wka_port_get
|                               open wka request
|                                              open response
|                              GSPN FC-GS
|                              RSPN FC-GS [NPIV-only]
|                              zfcp_fc_wka_port_put
|                               (--wka->refcount==0)
|                               sched delayed wka->work
|
~~~Case (1)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
zfcp_erp_wait
flush scan_work
|                                                      wka->work:
|                                                      wka->status=CLOSING
|                                                      close wka request
|                              scan_work:
|                              zfcp_fc_wka_port_get
|                               (wka->status==CLOSING)
|                               wka->status=OPENING
|                               open wka request
|                               wait_event
|                               |              close response
|                               |              wka->status=OFFLINE
|                               |              wake_up /*WRONG*/
~~~Case (2)~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|                                                      wka->work:
|                                                      wka->status=CLOSING
|                                                      close wka request
zfcp_erp_wait
flush scan_work
|                              scan_work:
|                              zfcp_fc_wka_port_get
|                               (wka->status==CLOSING)
|                               wka->status=OPENING
|                               open wka request
|                                              close response
|                                              wka->status=OFFLINE
|                                              wake_up /*WRONG&NOP*/
|                               wait_event /*NOP*/
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|                               (wka->status!=ONLINE)
|                               return -EIO
|                              return early
                                               open response
                                               wka->status=ONLINE
                                               wake_up /*NOP*/

So we erroneously end up with no automatic port scan. This is a big problem
when it happens during boot. The timing is influenced by v3.19 commit
18f87a67e6d6 ("zfcp: auto port scan resiliency").

Fix it by fully mutually excluding zfcp_fc_wka_port_get() and
zfcp_fc_wka_port_offline(). For that to work, we make the latter block
until we got the response for a close WKA port. In order not to penalize
the system workqueue, we move wka_port->work to our own adapter workqueue.
Note that before v2.6.30 commit 828bc1212a68 ("[SCSI] zfcp: Set WKA-port to
offline on adapter deactivation"), zfcp did block in
zfcp_fc_wka_port_offline() as well, but with a different condition.

While at it, make non-functional cleanups to improve code reading in
zfcp_fc_wka_port_get(). If we cannot send the WKA port open request, don't
rely on the subsequent wait_event condition to immediately let this case
pass without blocking. Also don't want to rely on the additional condition
handling the refcount to be skipped just to finally return with -EIO.

Link: https://lore.kernel.org/r/20220729162529.1620730-1-maier@linux.ibm.com
Fixes: 5ab944f97e09 ("[SCSI] zfcp: attach and release SAN nameserver port on demand")
Cc: <stable@vger.kernel.org> #v2.6.28+
Reviewed-by: Benjamin Block <bblock@linux.ibm.com>
Signed-off-by: Steffen Maier <maier@linux.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

scsi: ufs: core: Correct ufshcd_shutdown() flow

commit 00511d2abf5708ad05dd5d1c36adb2468d274698 upstream.

After ufshcd_wl_shutdown() set device power off and link off,
ufshcd_shutdown() could turn off clock/power. Also remove
pm_runtime_get_sync.

The reason why it is safe to remove pm_runtime_get_sync() is because:

 - ufshcd_wl_shutdown() -> pm_runtime_get_sync() will resume hba->dev too.

 - device resume(turn on clk/power) is not required, even if device is in
   RPM_SUSPENDED.

Link: https://lore.kernel.org/r/20220727030526.31022-1-peter.wang@mediatek.com
Fixes: b294ff3e3449 ("scsi: ufs: core: Enable power management for wlun")
Cc: <stable@vger.kernel.org> # 5.15.x
Reviewed-by: Stanley Chu <stanley.chu@mediatek.com>
Signed-off-by: Peter Wang <peter.wang@mediatek.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

video: fbdev: s3fb: Check the size of screen before memset_io()

[ Upstream commit 6ba592fa014f21f35a8ee8da4ca7b95a018f13e8 ]

In the function s3fb_set_par(), the value of 'screen_size' is
calculated by the user input. If the user provides the improper value,
the value of 'screen_size' may larger than 'info->screen_size', which
may cause the following bug:

[   54.083733] BUG: unable to handle page fault for address: ffffc90003000000
[   54.083742] #PF: supervisor write access in kernel mode
[   54.083744] #PF: error_code(0x0002) - not-present page
[   54.083760] RIP: 0010:memset_orig+0x33/0xb0
[   54.083782] Call Trace:
[   54.083788]  s3fb_set_par+0x1ec6/0x4040
[   54.083806]  fb_set_var+0x604/0xeb0
[   54.083836]  do_fb_ioctl+0x234/0x670

Fix the this by checking the value of 'screen_size' before memset_io().

Fixes: a268422de8bf ("fbdev driver for S3 Trio/Virge")
Signed-off-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

video: fbdev: arkfb: Check the size of screen before memset_io()

[ Upstream commit 96b550971c65d54d64728d8ba973487878a06454 ]

In the function arkfb_set_par(), the value of 'screen_size' is
calculated by the user input. If the user provides the improper value,
the value of 'screen_size' may larger than 'info->screen_size', which
may cause the following bug:

[  659.399066] BUG: unable to handle page fault for address: ffffc90003000000
[  659.399077] #PF: supervisor write access in kernel mode
[  659.399079] #PF: error_code(0x0002) - not-present page
[  659.399094] RIP: 0010:memset_orig+0x33/0xb0
[  659.399116] Call Trace:
[  659.399122]  arkfb_set_par+0x143f/0x24c0
[  659.399130]  fb_set_var+0x604/0xeb0
[  659.399161]  do_fb_ioctl+0x234/0x670
[  659.399189]  fb_ioctl+0xdd/0x130

Fix the this by checking the value of 'screen_size' before memset_io().

Fixes: 681e14730c73 ("arkfb: new framebuffer driver for ARK Logic cards")
Signed-off-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

video: fbdev: vt8623fb: Check the size of screen before memset_io()

[ Upstream commit ec0754c60217248fa77cc9005d66b2b55200ac06 ]

In the function vt8623fb_set_par(), the value of 'screen_size' is
calculated by the user input. If the user provides the improper value,
the value of 'screen_size' may larger than 'info->screen_size', which
may cause the following bug:

[  583.339036] BUG: unable to handle page fault for address: ffffc90005000000
[  583.339049] #PF: supervisor write access in kernel mode
[  583.339052] #PF: error_code(0x0002) - not-present page
[  583.339074] RIP: 0010:memset_orig+0x33/0xb0
[  583.339110] Call Trace:
[  583.339118]  vt8623fb_set_par+0x11cd/0x21e0
[  583.339146]  fb_set_var+0x604/0xeb0
[  583.339181]  do_fb_ioctl+0x234/0x670
[  583.339209]  fb_ioctl+0xdd/0x130

Fix the this by checking the value of 'screen_size' before memset_io().

Fixes: 558b7bd86c32 ("vt8623fb: new framebuffer driver for VIA VT8623")
Signed-off-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: do not allow to decompress files have FI_COMPRESS_RELEASED

[ Upstream commit 90be48bd9d29ece3965e5e8b21499b6db166e57b ]

If a file has FI_COMPRESS_RELEASED, all writes for it should not be
allowed. However, as of now, in case of compress_mode=user, writes
triggered by IOCTLs like F2FS_IOC_DE/COMPRESS_FILE are allowed unexpectly,
which could crash that file.
To fix it, let's do not allow F2FS_IOC_DE/COMPRESS_IOCTL if a file already
has FI_COMPRESS_RELEASED flag.

This is the reproduction process:
1.  $ touch ./file
2.  $ chattr +c ./file
3.  $ dd if=/dev/random of=./file bs=4096 count=30 conv=notrunc
4.  $ dd if=/dev/zero of=./file bs=4096 count=34 seek=30 conv=notrunc
5.  $ sync
6.  $ do_compress ./file      ; call F2FS_IOC_COMPRESS_FILE
7.  $ get_compr_blocks ./file ; call F2FS_IOC_GET_COMPRESS_BLOCKS
8.  $ release ./file          ; call F2FS_IOC_RELEASE_COMPRESS_BLOCKS
9.  $ do_compress ./file      ; call F2FS_IOC_COMPRESS_FILE again
10. $ get_compr_blocks ./file ; call F2FS_IOC_GET_COMPRESS_BLOCKS again

This reproduction process is tested in 128kb cluster size.
You can find compr_blocks has a negative value.

Fixes: 5fdb322ff2c2b ("f2fs: add F2FS_IOC_DECOMPRESS_FILE and F2FS_IOC_COMPRESS_FILE")

Signed-off-by: Junbeom Yeom <junbeom.yeom@samsung.com>
Signed-off-by: Sungjong Seo <sj1557.seo@samsung.com>
Signed-off-by: Youngjin Gil <youngjin.gil@samsung.com>
Signed-off-by: Jaewook Kim <jw5454.kim@samsung.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: allow compression for mmap files in compress_mode=user

[ Upstream commit 66d34fcbbe63ebd8584b792e0d741f6648100894 ]

Since commit e3c548323d32 ("f2fs: let's allow compression for mmap files"),
it has been allowed to compress mmap files. However, in compress_mode=user,
it is not allowed yet. To keep the same concept in both compress_modes,
f2fs_ioc_(de)compress_file() should also allow it.

Let's remove checking mmap files in f2fs_ioc_(de)compress_file() so that
the compression for mmap files is also allowed in compress_mode=user.

Signed-off-by: Sungjong Seo <sj1557.seo@samsung.com>
Reviewed-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

x86/entry: Build thunk_$(BITS) only if CONFIG_PREEMPTION=y

[ Upstream commit de979c83574abf6e78f3fa65b716515c91b2613d ]

With CONFIG_PREEMPTION disabled, arch/x86/entry/thunk_$(BITS).o becomes
an empty object file.

With some old versions of binutils (i.e., 2.35.90.20210113-1ubuntu1) the
GNU assembler doesn't generate a symbol table for empty object files and
objtool fails with the following error when a valid symbol table cannot
be found:

  arch/x86/entry/thunk_64.o: warning: objtool: missing symbol table

To prevent this from happening, build thunk_$(BITS).o only if
CONFIG_PREEMPTION is enabled.

BugLink: https://bugs.launchpad.net/bugs/1911359
Fixes: 320100a5ffe5 ("x86/entry: Remove the TRACE_IRQS cruft")
Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/Ys/Ke7EWjcX+ZlXO@arighi-desktop
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched/core: Do not requeue task on CPU excluded from cpus_mask

[ Upstream commit 751d4cbc43879229dbc124afefe240b70fd29a85 ]

The following warning was triggered on a large machine early in boot on
a distribution kernel but the same problem should also affect mainline.

   WARNING: CPU: 439 PID: 10 at ../kernel/workqueue.c:2231 process_one_work+0x4d/0x440
   Call Trace:
    <TASK>
    rescuer_thread+0x1f6/0x360
    kthread+0x156/0x180
    ret_from_fork+0x22/0x30
    </TASK>

Commit c6e7bd7afaeb ("sched/core: Optimize ttwu() spinning on p->on_cpu")
optimises ttwu by queueing a task that is descheduling on the wakelist,
but does not check if the task descheduling is still allowed to run on that CPU.

In this warning, the problematic task is a workqueue rescue thread which
checks if the rescue is for a per-cpu workqueue and running on the wrong CPU.
While this is early in boot and it should be possible to create workers,
the rescue thread may still used if the MAYDAY_INITIAL_TIMEOUT is reached
or MAYDAY_INTERVAL and on a sufficiently large machine, the rescue
thread is being used frequently.

Tracing confirmed that the task should have migrated properly using the
stopper thread to handle the migration. However, a parallel wakeup from udev
running on another CPU that does not share CPU cache observes p->on_cpu and
uses task_cpu(p), queues the task on the old CPU and triggers the warning.

Check that the wakee task that is descheduling is still allowed to run
on its current CPU and if not, wait for the descheduling to complete
and select an allowed CPU.

Fixes: c6e7bd7afaeb ("sched/core: Optimize ttwu() spinning on p->on_cpu")
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20220804092119.20137-1-mgorman@techsingularity.net
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched: Remove the limitation of WF_ON_CPU on wakelist if wakee cpu is idle

[ Upstream commit f3dd3f674555bd9455c5ae7fafce0696bd9931b3 ]

Wakelist can help avoid cache bouncing and offload the overhead of waker
cpu. So far, using wakelist within the same llc only happens on
WF_ON_CPU, and this limitation could be removed to further improve
wakeup performance.

The commit 518cd6234178 ("sched: Only queue remote wakeups when
crossing cache boundaries") disabled queuing tasks on wakelist when
the cpus share llc. This is because, at that time, the scheduler must
send IPIs to do ttwu_queue_wakelist. Nowadays, ttwu_queue_wakelist also
supports TIF_POLLING, so this is not a problem now when the wakee cpu is
in idle polling.

Benefits:
  Queuing the task on idle cpu can help improving performance on waker cpu
  and utilization on wakee cpu, and further improve locality because
  the wakee cpu can handle its own rq. This patch helps improving rt on
  our real java workloads where wakeup happens frequently.

  Consider the normal condition (CPU0 and CPU1 share same llc)
  Before this patch:

         CPU0                                       CPU1

    select_task_rq()                                idle
    rq_lock(CPU1->rq)
    enqueue_task(CPU1->rq)
    notify CPU1 (by sending IPI or CPU1 polling)

                                                    resched()

  After this patch:

         CPU0                                       CPU1

    select_task_rq()                                idle
    add to wakelist of CPU1
    notify CPU1 (by sending IPI or CPU1 polling)

                                                    rq_lock(CPU1->rq)
                                                    enqueue_task(CPU1->rq)
                                                    resched()

  We see CPU0 can finish its work earlier. It only needs to put task to
  wakelist and return.
  While CPU1 is idle, so let itself handle its own runqueue data.

This patch brings no difference about IPI.
  This patch only takes effect when the wakee cpu is:
  1) idle polling
  2) idle not polling

  For 1), there will be no IPI with or without this patch.

  For 2), there will always be an IPI before or after this patch.
  Before this patch: waker cpu will enqueue task and check preempt. Since
  "idle" will be sure to be preempted, waker cpu must send a resched IPI.
  After this patch: waker cpu will put the task to the wakelist of wakee
  cpu, and send an IPI.

Benchmark:
We've tested schbench, unixbench, and hachbench on both x86 and arm64.

On x86 (Intel Xeon Platinum 8269CY):
  schbench -m 2 -t 8

    Latency percentiles (usec)              before        after
        50.0000th:                             8            6
        75.0000th:                            10            7
        90.0000th:                            11            8
        95.0000th:                            12            8
        *99.0000th:                           13           10
        99.5000th:                            15           11
        99.9000th:                            18           14

  Unixbench with full threads (104)
                                            before        after
    Dhrystone 2 using register variables  3011862938    3009935994  -0.06%
    Double-Precision Whetstone              617119.3      617298.5   0.03%
    Execl Throughput                         27667.3       27627.3  -0.14%
    File Copy 1024 bufsize 2000 maxblocks   785871.4      784906.2  -0.12%
    File Copy 256 bufsize 500 maxblocks     210113.6      212635.4   1.20%
    File Copy 4096 bufsize 8000 maxblocks  2328862.2     2320529.1  -0.36%
    Pipe Throughput                      145535622.8   145323033.2  -0.15%
    Pipe-based Context Switching           3221686.4     3583975.4  11.25%
    Process Creation                        101347.1      103345.4   1.97%
    Shell Scripts (1 concurrent)            120193.5      123977.8   3.15%
    Shell Scripts (8 concurrent)             17233.4       17138.4  -0.55%
    System Call Overhead                   5300604.8     5312213.6   0.22%

  hackbench -g 1 -l 100000
                                            before        after
    Time                                     3.246        2.251

On arm64 (Ampere Altra):
  schbench -m 2 -t 8

    Latency percentiles (usec)              before        after
        50.0000th:                            14           10
        75.0000th:                            19           14
        90.0000th:                            22           16
        95.0000th:                            23           16
        *99.0000th:                           24           17
        99.5000th:                            24           17
        99.9000th:                            28           25

  Unixbench with full threads (80)
                                            before        after
    Dhrystone 2 using register variables  3536194249    3537019613   0.02%
    Double-Precision Whetstone              629383.6      629431.6   0.01%
    Execl Throughput                         65920.5       65846.2  -0.11%
    File Copy 1024 bufsize 2000 maxblocks  1063722.8     1064026.8   0.03%
    File Copy 256 bufsize 500 maxblocks     322684.5      318724.5  -1.23%
    File Copy 4096 bufsize 8000 maxblocks  2348285.3     2328804.8  -0.83%
    Pipe Throughput                      133542875.3   131619389.8  -1.44%
    Pipe-based Context Switching           3215356.1     3576945.1  11.25%
    Process Creation                        108520.5      120184.6  10.75%
    Shell Scripts (1 concurrent)            122636.3        121888  -0.61%
    Shell Scripts (8 concurrent)             17462.1       17381.4  -0.46%
    System Call Overhead                   4429998.9     4435006.7   0.11%

  hackbench -g 1 -l 100000
                                            before        after
    Time                                     4.217        2.916

Our patch has improvement on schbench, hackbench
and Pipe-based Context Switching of unixbench
when there exists idle cpus,
and no obvious regression on other tests of unixbench.
This can help improve rt in scenes where wakeup happens frequently.

Signed-off-by: Tianchen Ding <dtcccc@linux.alibaba.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Valentin Schneider <vschneid@redhat.com>
Link: https://lore.kernel.org/r/20220608233412.327341-3-dtcccc@linux.alibaba.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched: Fix the check of nr_running at queue wakelist

[ Upstream commit 28156108fecb1f808b21d216e8ea8f0d205a530c ]

The commit 2ebb17717550 ("sched/core: Offload wakee task activation if it
the wakee is descheduling") checked rq->nr_running <= 1 to avoid task
stacking when WF_ON_CPU.

Per the ordering of writes to p->on_rq and p->on_cpu, observing p->on_cpu
(WF_ON_CPU) in ttwu_queue_cond() implies !p->on_rq, IOW p has gone through
the deactivate_task() in __schedule(), thus p has been accounted out of
rq->nr_running. As such, the task being the only runnable task on the rq
implies reading rq->nr_running == 0 at that point.

The benchmark result is in [1].

[1] https://lore.kernel.org/all/e34de686-4e85-bde1-9f3c-9bbc86b38627@linux.alibaba.com/

Suggested-by: Valentin Schneider <vschneid@redhat.com>
Signed-off-by: Tianchen Ding <dtcccc@linux.alibaba.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Valentin Schneider <vschneid@redhat.com>
Link: https://lore.kernel.org/r/20220608233412.327341-2-dtcccc@linux.alibaba.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

tools/thermal: Fix possible path truncations

[ Upstream commit 6c58cf40e3a1d2f47c09d3489857e9476316788a ]

A build with -D_FORTIFY_SOURCE=2 enabled will produce the following warnings:

sysfs.c:63:30: warning: '%s' directive output may be truncated writing up to 255 bytes into a region of size between 0 and 255 [-Wformat-truncation=]
  snprintf(filepath, 256, "%s/%s", path, filename);
                              ^~
Bump up the buffer to PATH_MAX which is the limit and account for all of
the possible NUL and separators that could lead to exceeding the
allocated buffer sizes.

Fixes: 94f69966faf8 ("tools/thermal: Introduce tmon, a tool for thermal subsystem")
Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

video: fbdev: arkfb: Fix a divide-by-zero bug in ark_set_pixclock()

[ Upstream commit 2f1c4523f7a3aaabe7e53d3ebd378292947e95c8 ]

Since the user can control the arguments of the ioctl() from the user
space, under special arguments that may result in a divide-by-zero bug
in:
  drivers/video/fbdev/arkfb.c:784: ark_set_pixclock(info, (hdiv * info->var.pixclock) / hmul);
with hdiv=1, pixclock=1 and hmul=2 you end up with (1*1)/2 = (int) 0.
and then in:
  drivers/video/fbdev/arkfb.c:504: rv = dac_set_freq(par->dac, 0, 1000000000 / pixclock);
we'll get a division-by-zero.

The following log can reveal it:

divide error: 0000 [#1] PREEMPT SMP KASAN PTI
RIP: 0010:ark_set_pixclock drivers/video/fbdev/arkfb.c:504 [inline]
RIP: 0010:arkfb_set_par+0x10fc/0x24c0 drivers/video/fbdev/arkfb.c:784
Call Trace:
 fb_set_var+0x604/0xeb0 drivers/video/fbdev/core/fbmem.c:1034
 do_fb_ioctl+0x234/0x670 drivers/video/fbdev/core/fbmem.c:1110
 fb_ioctl+0xdd/0x130 drivers/video/fbdev/core/fbmem.c:1189

Fix this by checking the argument of ark_set_pixclock() first.

Fixes: 681e14730c73 ("arkfb: new framebuffer driver for ARK Logic cards")
Signed-off-by: Zheyu Ma <zheyuma97@gmail.com>
Signed-off-by: Helge Deller <deller@gmx.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

x86/numa: Use cpumask_available instead of hardcoded NULL check

[ Upstream commit 625395c4a0f4775e0fe00f616888d2e6c1ba49db ]

GCC-12 started triggering a new warning:

  arch/x86/mm/numa.c: In function ‘cpumask_of_node’:
  arch/x86/mm/numa.c:916:39: warning: the comparison will always evaluate as ‘false’ for the address of ‘node_to_cpumask_map’ will never be NULL [-Waddress]
    916 |         if (node_to_cpumask_map[node] == NULL) {
        |                                       ^~

node_to_cpumask_map is of type cpumask_var_t[].

When CONFIG_CPUMASK_OFFSTACK is set, cpumask_var_t is typedef'd to a
pointer for dynamic allocation, else to an array of one element. The
"wicked game" can be checked on line 700 of include/linux/cpumask.h.

The original code in debug_cpumask_set_cpu() and cpumask_of_node() were
probably written by the original authors with CONFIG_CPUMASK_OFFSTACK=y
(i.e. dynamic allocation) in mind, checking if the cpumask was available
via a direct NULL check.

When CONFIG_CPUMASK_OFFSTACK is not set, GCC gives the above warning
while compiling the kernel.

Fix that by using cpumask_available(), which does the NULL check when
CONFIG_CPUMASK_OFFSTACK is set, otherwise returns true. Use it wherever
such checks are made.

Conditional definitions of cpumask_available() can be found along with
the definition of cpumask_var_t. Check the cpumask.h reference mentioned
above.

Fixes: c032ef60d1aa ("cpumask: convert node_to_cpumask_map[] to cpumask_var_t")
Fixes: de2d9445f162 ("x86: Unify node_to_cpumask_map handling between 32 and 64bit")
Signed-off-by: Siddh Raman Pant <code@siddh.me>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20220731160913.632092-1-code@siddh.me
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched, cpuset: Fix dl_cpu_busy() panic due to empty cs->cpus_allowed

[ Upstream commit b6e8d40d43ae4dec00c8fea2593eeea3114b8f44 ]

With cgroup v2, the cpuset's cpus_allowed mask can be empty indicating
that the cpuset will just use the effective CPUs of its parent. So
cpuset_can_attach() can call task_can_attach() with an empty mask.
This can lead to cpumask_any_and() returns nr_cpu_ids causing the call
to dl_bw_of() to crash due to percpu value access of an out of bound
CPU value. For example:

[80468.182258] BUG: unable to handle page fault for address: ffffffff8b6648b0
  :
[80468.191019] RIP: 0010:dl_cpu_busy+0x30/0x2b0
  :
[80468.207946] Call Trace:
[80468.208947]  cpuset_can_attach+0xa0/0x140
[80468.209953]  cgroup_migrate_execute+0x8c/0x490
[80468.210931]  cgroup_update_dfl_csses+0x254/0x270
[80468.211898]  cgroup_subtree_control_write+0x322/0x400
[80468.212854]  kernfs_fop_write_iter+0x11c/0x1b0
[80468.213777]  new_sync_write+0x11f/0x1b0
[80468.214689]  vfs_write+0x1eb/0x280
[80468.215592]  ksys_write+0x5f/0xe0
[80468.216463]  do_syscall_64+0x5c/0x80
[80468.224287]  entry_SYSCALL_64_after_hwframe+0x44/0xae

Fix that by using effective_cpus instead. For cgroup v1, effective_cpus
is the same as cpus_allowed. For v2, effective_cpus is the real cpumask
to be used by tasks within the cpuset anyway.

Also update task_can_attach()'s 2nd argument name to cs_effective_cpus to
reflect the change. In addition, a check is added to task_can_attach()
to guard against the possibility that cpumask_any_and() may return a
value >= nr_cpu_ids.

Fixes: 7f51412a415d ("sched/deadline: Fix bandwidth check/update when migrating tasks between exclusive cpusets")
Signed-off-by: Waiman Long <longman@redhat.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Juri Lelli <juri.lelli@redhat.com>
Link: https://lore.kernel.org/r/20220803015451.2219567-1-longman@redhat.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

sched/deadline: Merge dl_task_can_attach() and dl_cpu_busy()

[ Upstream commit 772b6539fdda31462cc08368e78df60b31a58bab ]

Both functions are doing almost the same, that is checking if admission
control is still respected.

With exclusive cpusets, dl_task_can_attach() checks if the destination
cpuset (i.e. its root domain) has enough CPU capacity to accommodate the
task.
dl_cpu_busy() checks if there is enough CPU capacity in the cpuset in
case the CPU is hot-plugged out.

dl_task_can_attach() is used to check if a task can be admitted while
dl_cpu_busy() is used to check if a CPU can be hotplugged out.

Make dl_cpu_busy() able to deal with a task and use it instead of
dl_task_can_attach() in task_can_attach().

Signed-off-by: Dietmar Eggemann <dietmar.eggemann@arm.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Juri Lelli <juri.lelli@redhat.com>
Link: https://lore.kernel.org/r/20220302183433.333029-4-dietmar.eggemann@arm.com
Signed-off-by: Sasha Levin <sashal@kernel.org>

scripts/faddr2line: Fix vmlinux detection on arm64

[ Upstream commit b6a5068854cfe372da7dee3224dcf023ed5b00cb ]

Since commit dcea997beed6 ("faddr2line: Fix overlapping text section
failures, the sequel"), faddr2line is completely broken on arm64.

For some reason, on arm64, the vmlinux ELF object file type is ET_DYN
rather than ET_EXEC.  Check for both when determining whether the object
is vmlinux.

Modules and vmlinux.o have type ET_REL on all arches.

Fixes: dcea997beed6 ("faddr2line: Fix overlapping text section failures, the sequel")
Reported-by: John Garry <john.garry@huawei.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Tested-by: John Garry <john.garry@huawei.com>
Link: https://lore.kernel.org/r/dad1999737471b06d6188ce4cdb11329aa41682c.1658426357.git.jpoimboe@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>