shmem: fix smaps BUG sleeping while atomic

commit e5548f85b4527c4c803b7eae7887c10bf8f90c97 upstream.

smaps_pte_hole_lookup() is calling shmem_partial_swap_usage() with page
table lock held: but shmem_partial_swap_usage() does cond_resched_rcu() if
need_resched(): "BUG: sleeping function called from invalid context".

Since shmem_partial_swap_usage() is designed to count across a range, but
smaps_pte_hole_lookup() only calls it for a single page slot, just break
out of the loop on the last or only page, before checking need_resched().

Link: https://lkml.kernel.org/r/6fe3b3ec-abdf-332f-5c23-6a3b3a3b11a9@google.com
Fixes: 230100321518 ("mm/smaps: simplify shmem handling of pte holes")
Signed-off-by: Hugh Dickins <hughd@google.com>
Acked-by: Peter Xu <peterx@redhat.com>
Cc: <stable@vger.kernel.org> [5.16+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm,ima,kexec,of: use memblock_free_late from ima_free_kexec_buffer

commit f0362a253606e2031f8d61c74195d4d6556e12a4 upstream.

The code calling ima_free_kexec_buffer runs long after the memblock
allocator has already been torn down, potentially resulting in a use
after free in memblock_isolate_range.

With KASAN or KFENCE, this use after free will result in a BUG
from the idle task, and a subsequent kernel panic.

Switch ima_free_kexec_buffer over to memblock_free_late to avoid
that issue.

Fixes: fee3ff99bc67 ("powerpc: Move arch independent ima kexec functions to drivers/of/kexec.c")
Cc: stable@kernel.org
Signed-off-by: Rik van Riel <riel@surriel.com>
Suggested-by: Mike Rappoport <rppt@kernel.org>
Link: https://lore.kernel.org/r/20230817135759.0888e5ef@imladris.surriel.com
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: Fix slab-out-of-bounds error in devm_clk_release()

commit 66fbfb35da47f391bdadf9fa7ceb88af4faa9022 upstream.

Problem can be reproduced by unloading snd_soc_simple_card, because in
devm_get_clk_from_child() devres data is allocated as `struct clk`, but
devm_clk_release() expects devres data to be `struct devm_clk_state`.

KASAN report:
 ==================================================================
 BUG: KASAN: slab-out-of-bounds in devm_clk_release+0x20/0x54
 Read of size 8 at addr ffffff800ee09688 by task (udev-worker)/287

 Call trace:
  dump_backtrace+0xe8/0x11c
  show_stack+0x1c/0x30
  dump_stack_lvl+0x60/0x78
  print_report+0x150/0x450
  kasan_report+0xa8/0xf0
  __asan_load8+0x78/0xa0
  devm_clk_release+0x20/0x54
  release_nodes+0x84/0x120
  devres_release_all+0x144/0x210
  device_unbind_cleanup+0x1c/0xac
  really_probe+0x2f0/0x5b0
  __driver_probe_device+0xc0/0x1f0
  driver_probe_device+0x68/0x120
  __driver_attach+0x140/0x294
  bus_for_each_dev+0xec/0x160
  driver_attach+0x38/0x44
  bus_add_driver+0x24c/0x300
  driver_register+0xf0/0x210
  __platform_driver_register+0x48/0x54
  asoc_simple_card_init+0x24/0x1000 [snd_soc_simple_card]
  do_one_initcall+0xac/0x340
  do_init_module+0xd0/0x300
  load_module+0x2ba4/0x3100
  __do_sys_init_module+0x2c8/0x300
  __arm64_sys_init_module+0x48/0x5c
  invoke_syscall+0x64/0x190
  el0_svc_common.constprop.0+0x124/0x154
  do_el0_svc+0x44/0xdc
  el0_svc+0x14/0x50
  el0t_64_sync_handler+0xec/0x11c
  el0t_64_sync+0x14c/0x150

 Allocated by task 287:
  kasan_save_stack+0x38/0x60
  kasan_set_track+0x28/0x40
  kasan_save_alloc_info+0x20/0x30
  __kasan_kmalloc+0xac/0xb0
  __kmalloc_node_track_caller+0x6c/0x1c4
  __devres_alloc_node+0x44/0xb4
  devm_get_clk_from_child+0x44/0xa0
  asoc_simple_parse_clk+0x1b8/0x1dc [snd_soc_simple_card_utils]
  simple_parse_node.isra.0+0x1ec/0x230 [snd_soc_simple_card]
  simple_dai_link_of+0x1bc/0x334 [snd_soc_simple_card]
  __simple_for_each_link+0x2ec/0x320 [snd_soc_simple_card]
  asoc_simple_probe+0x468/0x4dc [snd_soc_simple_card]
  platform_probe+0x90/0xf0
  really_probe+0x118/0x5b0
  __driver_probe_device+0xc0/0x1f0
  driver_probe_device+0x68/0x120
  __driver_attach+0x140/0x294
  bus_for_each_dev+0xec/0x160
  driver_attach+0x38/0x44
  bus_add_driver+0x24c/0x300
  driver_register+0xf0/0x210
  __platform_driver_register+0x48/0x54
  asoc_simple_card_init+0x24/0x1000 [snd_soc_simple_card]
  do_one_initcall+0xac/0x340
  do_init_module+0xd0/0x300
  load_module+0x2ba4/0x3100
  __do_sys_init_module+0x2c8/0x300
  __arm64_sys_init_module+0x48/0x5c
  invoke_syscall+0x64/0x190
  el0_svc_common.constprop.0+0x124/0x154
  do_el0_svc+0x44/0xdc
  el0_svc+0x14/0x50
  el0t_64_sync_handler+0xec/0x11c
  el0t_64_sync+0x14c/0x150

 The buggy address belongs to the object at ffffff800ee09600
  which belongs to the cache kmalloc-256 of size 256
 The buggy address is located 136 bytes inside of
  256-byte region [ffffff800ee09600, ffffff800ee09700)

 The buggy address belongs to the physical page:
 page:000000002d97303b refcount:1 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x4ee08
 head:000000002d97303b order:1 compound_mapcount:0 compound_pincount:0
 flags: 0x10200(slab|head|zone=0)
 raw: 0000000000010200 0000000000000000 dead000000000122 ffffff8002c02480
 raw: 0000000000000000 0000000080100010 00000001ffffffff 0000000000000000
 page dumped because: kasan: bad access detected

 Memory state around the buggy address:
  ffffff800ee09580: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
  ffffff800ee09600: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
 >ffffff800ee09680: 00 fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
                       ^
  ffffff800ee09700: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
  ffffff800ee09780: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
 ==================================================================

Fixes: abae8e57e49a ("clk: generalize devm_clk_get() a bit")
Signed-off-by: Andrey Skvortsov <andrej.skvortzov@gmail.com>
Link: https://lore.kernel.org/r/20230805084847.3110586-1-andrej.skvortzov@gmail.com
Signed-off-by: Stephen Boyd <sboyd@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

NFSv4: Fix dropped lock for racing OPEN and delegation return

commit 1cbc11aaa01f80577b67ae02c73ee781112125fd upstream.

Commmit f5ea16137a3f ("NFSv4: Retry LOCK on OLD_STATEID during delegation
return") attempted to solve this problem by using nfs4's generic async error
handling, but introduced a regression where v4.0 lock recovery would hang.
The additional complexity introduced by overloading that error handling is
not necessary for this case.  This patch expects that commit to be
reverted.

The problem as originally explained in the above commit is:

    There's a small window where a LOCK sent during a delegation return can
    race with another OPEN on client, but the open stateid has not yet been
    updated.  In this case, the client doesn't handle the OLD_STATEID error
    from the server and will lose this lock, emitting:
    "NFS: nfs4_handle_delegation_recall_error: unhandled error -10024".

Fix this by using the old_stateid refresh helpers if the server replies
with OLD_STATEID.

Suggested-by: Trond Myklebust <trondmy@hammerspace.com>
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

platform/x86: ideapad-laptop: Add support for new hotkeys found on ThinkBook 14s Yoga ITL

commit a260f7d726fde52c0278bd3fa085a758639bcee2 upstream.

The Lenovo Thinkbook 14s Yoga ITL has 4 new symbols/shortcuts on their
F9-F11 and PrtSc keys:

F9:    Has a symbol of a head with a headset, the manual says "Service key"
F10:   Has a symbol of a telephone horn which has been picked up from the
       receiver, the manual says: "Answer incoming calls"
F11:   Has a symbol of a telephone horn which is resting on the receiver,
       the manual says: "Reject incoming calls"
PrtSc: Has a symbol of a siccor and a dashed ellipse, the manual says:
       "Open the Windows 'Snipping' Tool app"

This commit adds support for these 4 new hkey events.

Signed-off-by: André Apitzsch <git@apitzsch.eu>
Link: https://lore.kernel.org/r/20230819-lenovo_keys-v1-1-9d34eac88e0a@apitzsch.eu
Reviewed-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: mac80211: limit reorder_buf_filtered to avoid UBSAN warning

commit b98c16107cc1647242abbd11f234c05a3a5864f6 upstream.

The commit 06470f7468c8 ("mac80211: add API to allow filtering frames in BA sessions")
added reorder_buf_filtered to mark frames filtered by firmware, and it
can only work correctly if hw.max_rx_aggregation_subframes <= 64 since
it stores the bitmap in a u64 variable.

However, new HE or EHT devices can support BlockAck number up to 256 or
1024, and then using a higher subframe index leads UBSAN warning:

 UBSAN: shift-out-of-bounds in net/mac80211/rx.c:1129:39
 shift exponent 215 is too large for 64-bit type 'long long unsigned int'
 Call Trace:
  <IRQ>
  dump_stack_lvl+0x48/0x70
  dump_stack+0x10/0x20
  __ubsan_handle_shift_out_of_bounds+0x1ac/0x360
  ieee80211_release_reorder_frame.constprop.0.cold+0x64/0x69 [mac80211]
  ieee80211_sta_reorder_release+0x9c/0x400 [mac80211]
  ieee80211_prepare_and_rx_handle+0x1234/0x1420 [mac80211]
  ieee80211_rx_list+0xaef/0xf60 [mac80211]
  ieee80211_rx_napi+0x53/0xd0 [mac80211]

Since only old hardware that supports <=64 BlockAck uses
ieee80211_mark_rx_ba_filtered_frames(), limit the use as it is, so add a
WARN_ONCE() and comment to note to avoid using this function if hardware
capability is not suitable.

Signed-off-by: Ping-Ke Shih <pkshih@realtek.com>
Link: https://lore.kernel.org/r/20230818014004.16177-1-pkshih@realtek.com
[edit commit message]
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ibmveth: Use dcbf rather than dcbfl

commit bfedba3b2c7793ce127680bc8f70711e05ec7a17 upstream.

When building for power4, newer binutils don't recognise the "dcbfl"
extended mnemonic.

dcbfl RA, RB is equivalent to dcbf RA, RB, 1.

Switch to "dcbf" to avoid the build error.

Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ASoC: cs35l41: Correct amp_gain_tlv values

commit 1613781d7e8a93618ff3a6b37f81f06769b53717 upstream.

The current analog gain TLV seems to have completely incorrect values in
it. The gain starts at 0.5dB, proceeds in 1dB steps, and has no mute
value, correct the control to match.

Signed-off-by: Charles Keepax <ckeepax@opensource.cirrus.com>
Link: https://lore.kernel.org/r/20230823085308.753572-1-ckeepax@opensource.cirrus.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ASoC: amd: yc: Add VivoBook Pro 15 to quirks list for acp6x

commit 3b1f08833c45d0167741e4097b0150e7cf086102 upstream.

VivoBook Pro 15 Ryzen Edition uses Ryzen 6800H processor, and adding to
 quirks list for acp6x will enable internal mic.

Signed-off-by: BrenoRCBrito <brenorcbrito@gmail.com>
Link: https://lore.kernel.org/r/20230818211417.32167-1-brenorcbrito@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring/msg_ring: fix missing lock on overflow for IOPOLL

Commit e12d7a46f65ae4b7d58a5e0c1cbfa825cf8d830d upstream.

If the target ring is configured with IOPOLL, then we always need to hold
the target ring uring_lock before posting CQEs. We could just grab it
unconditionally, but since we don't expect many target rings to be of this
type, make grabbing the uring_lock conditional on the ring type.

Link: https://lore.kernel.org/io-uring/Y8krlYa52%2F0YGqkg@ip-172-31-85-199.ec2.internal/
Reported-by: Xingyuan Mo <hdthky0@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring/msg_ring: move double lock/unlock helpers higher up

Commit 423d5081d0451faa59a707e57373801da5b40141 upstream.

In preparation for needing them somewhere else, move them and get rid of
the unused 'issue_flags' for the unlock side.

No functional changes in this patch.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring: extract a io_msg_install_complete helper

Commit 172113101641cf1f9628c528ec790cb809f2b704 upstream.

Extract a helper called io_msg_install_complete() from io_msg_send_fd(),
will be used later.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/1500ca1054cc4286a3ee1c60aacead57fcdfa02a.1670384893.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring: get rid of double locking

Commit 11373026f2960390d5e330df4e92735c4265c440 upstream.

We don't need to take both uring_locks at once, msg_ring can be split in
two parts, first getting a file from the filetable of the first ring and
then installing it into the second one.

Signed-off-by: Pavel Begunkov <asml.silence@gmail.com>
Link: https://lore.kernel.org/r/a80ecc2bc99c3b3f2cf20015d618b7c51419a797.1670384893.git.asml.silence@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: x86/mmu: Fix an sign-extension bug with mmu_seq that hangs vCPUs

Upstream commit ba6e3fe25543 ("KVM: x86/mmu: Grab mmu_invalidate_seq in
kvm_faultin_pfn()") unknowingly fixed the bug in v6.3 when refactoring
how KVM tracks the sequence counter snapshot.

Take the vCPU's mmu_seq snapshot as an "unsigned long" instead of an "int"
when checking to see if a page fault is stale, as the sequence count is
stored as an "unsigned long" everywhere else in KVM.  This fixes a bug
where KVM will effectively hang vCPUs due to always thinking page faults
are stale, which results in KVM refusing to "fix" faults.

mmu_invalidate_seq (née mmu_notifier_seq) is a sequence counter used when
KVM is handling page faults to detect if userspace mappings relevant to
the guest were invalidated between snapshotting the counter and acquiring
mmu_lock, i.e. to ensure that the userspace mapping KVM is using to
resolve the page fault is fresh.  If KVM sees that the counter has
changed, KVM simply resumes the guest without fixing the fault.

What _should_ happen is that the source of the mmu_notifier invalidations
eventually goes away, mmu_invalidate_seq becomes stable, and KVM can once
again fix guest page fault(s).

But for a long-lived VM and/or a VM that the host just doesn't particularly
like, it's possible for a VM to be on the receiving end of 2 billion (with
a B) mmu_notifier invalidations.  When that happens, bit 31 will be set in
mmu_invalidate_seq.  This causes the value to be turned into a 32-bit
negative value when implicitly cast to an "int" by is_page_fault_stale(),
and then sign-extended into a 64-bit unsigned when the signed "int" is
implicitly cast back to an "unsigned long" on the call to
mmu_invalidate_retry_hva().

As a result of the casting and sign-extension, given a sequence counter of
e.g. 0x8002dc25, mmu_invalidate_retry_hva() ends up doing

if (0x8002dc25 != 0xffffffff8002dc25)

and signals that the page fault is stale and needs to be retried even
though the sequence counter is stable, and KVM effectively hangs any vCPU
that takes a page fault (EPT violation or #NPF when TDP is enabled).

Reported-by: Brian Rak <brak@vultr.com>
Reported-by: Amaan Cheval <amaan.cheval@gmail.com>
Reported-by: Eric Wheeler <kvm@lists.ewheeler.net>
Closes: https://lore.kernel.org/all/f023d927-52aa-7e08-2ee5-59a2fbc65953@gameservers.com
Fixes: a955cad84cda ("KVM: x86/mmu: Retry page fault if root is invalidated by memslot update")
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: x86: Preserve TDP MMU roots until they are explicitly invalidated

commit edbdb43fc96b11b3bfa531be306a1993d9fe89ec upstream.

Preserve TDP MMU roots until they are explicitly invalidated by gifting
the TDP MMU itself a reference to a root when it is allocated.  Keeping a
reference in the TDP MMU fixes a flaw where the TDP MMU exhibits terrible
performance, and can potentially even soft-hang a vCPU, if a vCPU
frequently unloads its roots, e.g. when KVM is emulating SMI+RSM.

When KVM emulates something that invalidates _all_ TLB entries, e.g. SMI
and RSM, KVM unloads all of the vCPUs roots (KVM keeps a small per-vCPU
cache of previous roots).  Unloading roots is a simple way to ensure KVM
flushes and synchronizes all roots for the vCPU, as KVM flushes and syncs
when allocating a "new" root (from the vCPU's perspective).

In the shadow MMU, KVM keeps track of all shadow pages, roots included, in
a per-VM hash table.  Unloading a shadow MMU root just wipes it from the
per-vCPU cache; the root is still tracked in the per-VM hash table.  When
KVM loads a "new" root for the vCPU, KVM will find the old, unloaded root
in the per-VM hash table.

Unlike the shadow MMU, the TDP MMU doesn't track "inactive" roots in a
per-VM structure, where "active" in this case means a root is either
in-use or cached as a previous root by at least one vCPU.  When a TDP MMU
root becomes inactive, i.e. the last vCPU reference to the root is put,
KVM immediately frees the root (asterisk on "immediately" as the actual
freeing may be done by a worker, but for all intents and purposes the root
is gone).

The TDP MMU behavior is especially problematic for 1-vCPU setups, as
unloading all roots effectively frees all roots.  The issue is mitigated
to some degree in multi-vCPU setups as a different vCPU usually holds a
reference to an unloaded root and thus keeps the root alive, allowing the
vCPU to reuse its old root after unloading (with a flush+sync).

The TDP MMU flaw has been known for some time, as until very recently,
KVM's handling of CR0.WP also triggered unloading of all roots.  The
CR0.WP toggling scenario was eventually addressed by not unloading roots
when _only_ CR0.WP is toggled, but such an approach doesn't Just Work
for emulating SMM as KVM must emulate a full TLB flush on entry and exit
to/from SMM.  Given that the shadow MMU plays nice with unloading roots
at will, teaching the TDP MMU to do the same is far less complex than
modifying KVM to track which roots need to be flushed before reuse.

Note, preserving all possible TDP MMU roots is not a concern with respect
to memory consumption.  Now that the role for direct MMUs doesn't include
information about the guest, e.g. CR0.PG, CR0.WP, CR4.SMEP, etc., there
are _at most_ six possible roots (where "guest_mode" here means L2):

  1. 4-level !SMM !guest_mode
  2. 4-level  SMM !guest_mode
  3. 5-level !SMM !guest_mode
  4. 5-level  SMM !guest_mode
  5. 4-level !SMM guest_mode
  6. 5-level !SMM guest_mode

And because each vCPU can track 4 valid roots, a VM can already have all
6 root combinations live at any given time.  Not to mention that, in
practice, no sane VMM will advertise different guest.MAXPHYADDR values
across vCPUs, i.e. KVM won't ever use both 4-level and 5-level roots for
a single VM.  Furthermore, the vast majority of modern hypervisors will
utilize EPT/NPT when available, thus the guest_mode=%true cases are also
unlikely to be utilized.

Reported-by: Jeremi Piotrowski <jpiotrowski@linux.microsoft.com>
Link: https://lore.kernel.org/all/959c5bce-beb5-b463-7158-33fc4a4f910c@linux.microsoft.com
Link: https://lkml.kernel.org/r/20220209170020.1775368-1-pbonzini%40redhat.com
Link: https://lore.kernel.org/all/20230322013731.102955-1-minipli@grsecurity.net
Link: https://lore.kernel.org/all/000000000000a0bc2b05f9dd7fab@google.com
Link: https://lore.kernel.org/all/000000000000eca0b905fa0f7756@google.com
Cc: Ben Gardon <bgardon@google.com>
Cc: David Matlack <dmatlack@google.com>
Cc: stable@vger.kernel.org
Tested-by: Jeremi Piotrowski <jpiotrowski@linux.microsoft.com>
Link: https://lore.kernel.org/r/20230426220323.3079789-1-seanjc@google.com
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bonding: fix macvlan over alb bond support

[ Upstream commit e74216b8def3803e98ae536de78733e9d7f3b109 ]

The commit 14af9963ba1e ("bonding: Support macvlans on top of tlb/rlb mode
bonds") aims to enable the use of macvlans on top of rlb bond mode. However,
the current rlb bond mode only handles ARP packets to update remote neighbor
entries. This causes an issue when a macvlan is on top of the bond, and
remote devices send packets to the macvlan using the bond's MAC address
as the destination. After delivering the packets to the macvlan, the macvlan
will rejects them as the MAC address is incorrect. Consequently, this commit
makes macvlan over bond non-functional.

To address this problem, one potential solution is to check for the presence
of a macvlan port on the bond device using netif_is_macvlan_port(bond->dev)
and return NULL in the rlb_arp_xmit() function. However, this approach
doesn't fully resolve the situation when a VLAN exists between the bond and
macvlan.

So let's just do a partial revert for commit 14af9963ba1e in rlb_arp_xmit().
As the comment said, Don't modify or load balance ARPs that do not originate
locally.

Fixes: 14af9963ba1e ("bonding: Support macvlans on top of tlb/rlb mode bonds")
Reported-by: susan.zheng@veritas.com
Closes: https://bugzilla.redhat.com/show_bug.cgi?id=2117816
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Acked-by: Jay Vosburgh <jay.vosburgh@canonical.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

rtnetlink: Reject negative ifindexes in RTM_NEWLINK

[ Upstream commit 30188bd7838c16a98a520db1fe9df01ffc6ed368 ]

Negative ifindexes are illegal, but the kernel does not validate the
ifindex in the ancillary header of RTM_NEWLINK messages, resulting in
the kernel generating a warning [1] when such an ifindex is specified.

Fix by rejecting negative ifindexes.

[1]
WARNING: CPU: 0 PID: 5031 at net/core/dev.c:9593 dev_index_reserve+0x1a2/0x1c0 net/core/dev.c:9593
[...]
Call Trace:
 <TASK>
 register_netdevice+0x69a/0x1490 net/core/dev.c:10081
 br_dev_newlink+0x27/0x110 net/bridge/br_netlink.c:1552
 rtnl_newlink_create net/core/rtnetlink.c:3471 [inline]
 __rtnl_newlink+0x115e/0x18c0 net/core/rtnetlink.c:3688
 rtnl_newlink+0x67/0xa0 net/core/rtnetlink.c:3701
 rtnetlink_rcv_msg+0x439/0xd30 net/core/rtnetlink.c:6427
 netlink_rcv_skb+0x16b/0x440 net/netlink/af_netlink.c:2545
 netlink_unicast_kernel net/netlink/af_netlink.c:1342 [inline]
 netlink_unicast+0x536/0x810 net/netlink/af_netlink.c:1368
 netlink_sendmsg+0x93c/0xe40 net/netlink/af_netlink.c:1910
 sock_sendmsg_nosec net/socket.c:728 [inline]
 sock_sendmsg+0xd9/0x180 net/socket.c:751
 ____sys_sendmsg+0x6ac/0x940 net/socket.c:2538
 ___sys_sendmsg+0x135/0x1d0 net/socket.c:2592
 __sys_sendmsg+0x117/0x1e0 net/socket.c:2621
 do_syscall_x64 arch/x86/entry/common.c:50 [inline]
 do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80
 entry_SYSCALL_64_after_hwframe+0x63/0xcd

Fixes: 38f7b870d4a6 ("[RTNETLINK]: Link creation API")
Reported-by: syzbot+5ba06978f34abb058571@syzkaller.appspotmail.com
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Link: https://lore.kernel.org/r/20230823064348.2252280-1-idosch@nvidia.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nf_tables: fix out of memory error handling

[ Upstream commit 5e1be4cdc98c989d5387ce94ff15b5ad06a5b681 ]

Several instances of pipapo_resize() don't propagate allocation failures,
this causes a crash when fault injection is enabled for gfp_kernel slabs.

Fixes: 3c4287f62044 ("nf_tables: Add set type for arbitrary concatenation of ranges")
Signed-off-by: Florian Westphal <fw@strlen.de>
Reviewed-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

netfilter: nf_tables: flush pending destroy work before netlink notifier

[ Upstream commit 2c9f0293280e258606e54ed2b96fa71498432eae ]

Destroy work waits for the RCU grace period then it releases the objects
with no mutex held. All releases objects follow this path for
transactions, therefore, order is guaranteed and references to top-level
objects in the hierarchy remain valid.

However, netlink notifier might interfer with pending destroy work.
rcu_barrier() is not correct because objects are not release via RCU
callback. Flush destroy work before releasing objects from netlink
notifier path.

Fixes: d4bc8271db21 ("netfilter: nf_tables: netlink notifier might race to release objects")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

i40e: fix potential NULL pointer dereferencing of pf->vf i40e_sync_vsi_filters()

[ Upstream commit 9525a3c38accd2e186f52443e35e633e296cc7f5 ]

Add check for pf->vf not being NULL before dereferencing
pf->vf[vsi->vf_id] in updating VSI filter sync.
Add a similar check before dereferencing !pf->vf[vsi->vf_id].trusted
in the condition for clearing promisc mode bit.

Fixes: c87c938f62d8 ("i40e: Add VF VLAN pruning")
Signed-off-by: Andrii Staikov <andrii.staikov@intel.com>
Signed-off-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: fix a qdisc modification with ambiguous command request

[ Upstream commit da71714e359b64bd7aab3bd56ec53f307f058133 ]

When replacing an existing root qdisc, with one that is of the same kind, the
request boils down to essentially a parameterization change  i.e not one that
requires allocation and grafting of a new qdisc. syzbot was able to create a
scenario which resulted in a taprio qdisc replacing an existing taprio qdisc
with a combination of NLM_F_CREATE, NLM_F_REPLACE and NLM_F_EXCL leading to
create and graft scenario.
The fix ensures that only when the qdisc kinds are different that we should
allow a create and graft, otherwise it goes into the "change" codepath.

While at it, fix the code and comments to improve readability.

While syzbot was able to create the issue, it did not zone on the root cause.
Analysis from Vladimir Oltean <vladimir.oltean@nxp.com> helped narrow it down.

v1->V2 changes:
- remove "inline" function definition (Vladmir)
- remove extrenous braces in branches (Vladmir)
- change inline function names (Pedro)
- Run tdc tests (Victor)
v2->v3 changes:
- dont break else/if (Simon)

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: syzbot+a3618a167af2021433cd@syzkaller.appspotmail.com
Closes: https://lore.kernel.org/netdev/20230816225759.g25x76kmgzya2gei@skbuf/T/
Tested-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Tested-by: Victor Nogueira <victor@mojatatu.com>
Reviewed-by: Pedro Tammela <pctammela@mojatatu.com>
Reviewed-by: Victor Nogueira <victor@mojatatu.com>
Signed-off-by: Jamal Hadi Salim <jhs@mojatatu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

igc: Fix the typo in the PTM Control macro

[ Upstream commit de43975721b97283d5f17eea4228faddf08f2681 ]

The IGC_PTM_CTRL_SHRT_CYC defines the time between two consecutive PTM
requests. The bit resolution of this field is six bits. That bit five was
missing in the mask. This patch comes to correct the typo in the
IGC_PTM_CTRL_SHRT_CYC macro.

Fixes: a90ec8483732 ("igc: Add support for PTP getcrosststamp()")
Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Link: https://lore.kernel.org/r/20230821171721.2203572-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

igb: Avoid starting unnecessary workqueues

[ Upstream commit b888c510f7b3d64ca75fc0f43b4a4bd1a611312f ]

If ptp_clock_register() fails or CONFIG_PTP isn't enabled, avoid starting
PTP related workqueues.

In this way we can fix this:
 BUG: unable to handle page fault for address: ffffc9000440b6f8
 #PF: supervisor read access in kernel mode
 #PF: error_code(0x0000) - not-present page
 PGD 100000067 P4D 100000067 PUD 1001e0067 PMD 107dc5067 PTE 0
 Oops: 0000 [#1] PREEMPT SMP
 [...]
 Workqueue: events igb_ptp_overflow_check
 RIP: 0010:igb_rd32+0x1f/0x60
 [...]
 Call Trace:
  igb_ptp_read_82580+0x20/0x50
  timecounter_read+0x15/0x60
  igb_ptp_overflow_check+0x1a/0x50
  process_one_work+0x1cb/0x3c0
  worker_thread+0x53/0x3f0
  ? rescuer_thread+0x370/0x370
  kthread+0x142/0x160
  ? kthread_associate_blkcg+0xc0/0xc0
  ret_from_fork+0x1f/0x30

Fixes: 1f6e8178d685 ("igb: Prevent dropped Tx timestamps via work items and interrupts.")
Fixes: d339b1331616 ("igb: add PTP Hardware Clock code")
Signed-off-by: Alessio Igor Bogani <alessio.bogani@elettra.eu>
Tested-by: Arpana Arland <arpanax.arland@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20230821171927.2203644-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

can: isotp: fix support for transmission of SF without flow control

[ Upstream commit 0bfe71159230bab79ee230225ae12ffecbb69f3e ]

The original implementation had a very simple handling for single frame
transmissions as it just sent the single frame without a timeout handling.

With the new echo frame handling the echo frame was also introduced for
single frames but the former exception ('simple without timers') has been
maintained by accident. This leads to a 1 second timeout when closing the
socket and to an -ECOMM error when CAN_ISOTP_WAIT_TX_DONE is selected.

As the echo handling is always active (also for single frames) remove the
wrong extra condition for single frames.

Fixes: 9f39d36530e5 ("can: isotp: add support for transmission without flow control")
Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Link: https://lore.kernel.org/r/20230821144547.6658-2-socketcan@hartkopp.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: bonding: do not set port down before adding to bond

[ Upstream commit be809424659c2844a2d7ab653aacca4898538023 ]

Before adding a port to bond, it need to be set down first. In the
lacpdu test the author set the port down specifically. But commit
a4abfa627c38 ("net: rtnetlink: Enslave device before bringing it up")
changed the operation order, the kernel will set the port down _after_
adding to bond. So all the ports will be down at last and the test failed.

In fact, the veth interfaces are already inactive when added. This
means there's no need to set them down again before adding to the bond.
Let's just remove the link down operation.

Fixes: a4abfa627c38 ("net: rtnetlink: Enslave device before bringing it up")
Reported-by: Zhengchao Shao <shaozhengchao@huawei.com>
Closes: https://lore.kernel.org/netdev/a0ef07c7-91b0-94bd-240d-944a330fcabd@huawei.com/
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Link: https://lore.kernel.org/r/20230817082459.1685972-1-liuhangbin@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ice: Fix NULL pointer deref during VF reset

[ Upstream commit 67f6317dfa609846a227a706532439a22828c24b ]

During stress test with attaching and detaching VF from KVM and
simultaneously changing VFs spoofcheck and trust there was a
NULL pointer dereference in ice_reset_vf that VF's VSI is null.

More than one instance of ice_reset_vf() can be running at a given
time. When we rebuild the VSI in ice_reset_vf, another reset can be
triaged from ice_service_task. In this case we can access the currently
uninitialized VSI and cause panic. The window for this racing condition
has been around for a long time but it's much worse after commit
227bf4500aaa ("ice: move VSI delete outside deconfig") because
the reset runs faster. ice_reset_vf() using vf->cfg_lock and when
we move this lock before accessing to the VF VSI, we can fix
BUG for all cases.

Panic occurs sometimes in ice_vsi_is_rx_queue_active() and sometimes
in ice_vsi_stop_all_rx_rings()

With our reproducer, we can hit BUG:
~8h before commit 227bf4500aaa ("ice: move VSI delete outside deconfig").
~20m after commit 227bf4500aaa ("ice: move VSI delete outside deconfig").
After this fix we are not able to reproduce it after ~48h

There was commit cf90b74341ee ("ice: Fix call trace with null VSI during
VF reset") which also tried to fix this issue, but it was only
partially resolved and the bug still exists.

[ 6420.658415] BUG: kernel NULL pointer dereference, address: 0000000000000000
[ 6420.665382] #PF: supervisor read access in kernel mode
[ 6420.670521] #PF: error_code(0x0000) - not-present page
[ 6420.675659] PGD 0
[ 6420.677679] Oops: 0000 [#1] PREEMPT SMP NOPTI
[ 6420.682038] CPU: 53 PID: 326472 Comm: kworker/53:0 Kdump: loaded Not tainted 5.14.0-317.el9.x86_64 #1
[ 6420.691250] Hardware name: Dell Inc. PowerEdge R750/04V528, BIOS 1.6.5 04/15/2022
[ 6420.698729] Workqueue: ice ice_service_task [ice]
[ 6420.703462] RIP: 0010:ice_vsi_is_rx_queue_active+0x2d/0x60 [ice]
[ 6420.705860] ice 0000:ca:00.0: VF 0 is now untrusted
[ 6420.709494] Code: 00 00 66 83 bf 76 04 00 00 00 48 8b 77 10 74 3e 31 c0 eb 0f 0f b7 97 76 04 00 00 48 83 c0 01 39 c2 7e 2b 48 8b 97 68 04 00 00 <0f> b7 0c 42 48 8b 96 20 13 00 00 48 8d 94 8a 00 00 12 00 8b 12 83
[ 6420.714426] ice 0000:ca:00.0 ens7f0: Setting MAC 22:22:22:22:22:00 on VF 0. VF driver will be reinitialized
[ 6420.733120] RSP: 0018:ff778d2ff383fdd8 EFLAGS: 00010246
[ 6420.733123] RAX: 0000000000000000 RBX: ff2acf1916294000 RCX: 0000000000000000
[ 6420.733125] RDX: 0000000000000000 RSI: ff2acf1f2c6401a0 RDI: ff2acf1a27301828
[ 6420.762346] RBP: ff2acf1a27301828 R08: 0000000000000010 R09: 0000000000001000
[ 6420.769476] R10: ff2acf1916286000 R11: 00000000019eba3f R12: ff2acf19066460d0
[ 6420.776611] R13: ff2acf1f2c6401a0 R14: ff2acf1f2c6401a0 R15: 00000000ffffffff
[ 6420.783742] FS:  0000000000000000(0000) GS:ff2acf28ffa80000(0000) knlGS:0000000000000000
[ 6420.791829] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[ 6420.797575] CR2: 0000000000000000 CR3: 00000016ad410003 CR4: 0000000000773ee0
[ 6420.804708] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[ 6420.811034] vfio-pci 0000:ca:01.0: enabling device (0000 -> 0002)
[ 6420.811840] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[ 6420.811841] PKRU: 55555554
[ 6420.811842] Call Trace:
[ 6420.811843]  <TASK>
[ 6420.811844]  ice_reset_vf+0x9a/0x450 [ice]
[ 6420.811876]  ice_process_vflr_event+0x8f/0xc0 [ice]
[ 6420.841343]  ice_service_task+0x23b/0x600 [ice]
[ 6420.845884]  ? __schedule+0x212/0x550
[ 6420.849550]  process_one_work+0x1e2/0x3b0
[ 6420.853563]  ? rescuer_thread+0x390/0x390
[ 6420.857577]  worker_thread+0x50/0x3a0
[ 6420.861242]  ? rescuer_thread+0x390/0x390
[ 6420.865253]  kthread+0xdd/0x100
[ 6420.868400]  ? kthread_complete_and_exit+0x20/0x20
[ 6420.873194]  ret_from_fork+0x1f/0x30
[ 6420.876774]  </TASK>
[ 6420.878967] Modules linked in: vfio_pci vfio_pci_core vfio_iommu_type1 vfio iavf vhost_net vhost vhost_iotlb tap tun xt_CHECKSUM xt_MASQUERADE xt_conntrack ipt_REJECT nf_reject_ipv4 nft_compat nft_chain_nat nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 nft_counter nf_tables bridge stp llc sctp ip6_udp_tunnel udp_tunnel nfp tls nfnetlink bluetooth mlx4_en mlx4_core rpcsec_gss_krb5 auth_rpcgss nfsv4 dns_resolver nfs lockd grace fscache netfs rfkill sunrpc intel_rapl_msr intel_rapl_common i10nm_edac nfit libnvdimm ipmi_ssif x86_pkg_temp_thermal intel_powerclamp coretemp irdma kvm_intel i40e kvm iTCO_wdt dcdbas ib_uverbs irqbypass iTCO_vendor_support mgag200 mei_me ib_core dell_smbios isst_if_mmio isst_if_mbox_pci rapl i2c_algo_bit drm_shmem_helper intel_cstate drm_kms_helper syscopyarea sysfillrect isst_if_common sysimgblt intel_uncore fb_sys_fops dell_wmi_descriptor wmi_bmof intel_vsec mei i2c_i801 acpi_ipmi ipmi_si i2c_smbus ipmi_devintf intel_pch_thermal acpi_power_meter pcspk
 r

Fixes: efe41860008e ("ice: Fix memory corruption in VF driver")
Fixes: f23df5220d2b ("ice: Fix spurious interrupt during removal of trusted VF")
Signed-off-by: Petr Oros <poros@redhat.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Revert "ice: Fix ice VF reset during iavf initialization"

[ Upstream commit 0ecff05e6c59dd82dbcb9706db911f7fd9f40fb8 ]

This reverts commit 7255355a0636b4eff08d5e8139c77d98f151c4fc.

After this commit we are not able to attach VF to VM:
virsh attach-interface v0 hostdev --managed 0000:41:01.0 --mac 52:52:52:52:52:52
error: Failed to attach interface
error: Cannot set interface MAC to 52:52:52:52:52:52 for ifname enp65s0f0np0 vf 0: Resource temporarily unavailable

ice_check_vf_ready_for_cfg() already contain waiting for reset.
New condition in ice_check_vf_ready_for_reset() causing only problems.

Fixes: 7255355a0636 ("ice: Fix ice VF reset during iavf initialization")
Signed-off-by: Petr Oros <poros@redhat.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Tested-by: Rafal Romanowski <rafal.romanowski@intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ice: fix receive buffer size miscalculation

[ Upstream commit 10083aef784031fa9f06c19a1b182e6fad5338d9 ]

The driver is misconfiguring the hardware for some values of MTU such that
it could use multiple descriptors to receive a packet when it could have
simply used one.

Change the driver to use a round-up instead of the result of a shift, as
the shift can truncate the lower bits of the size, and result in the
problem noted above. It also aligns this driver with similar code in i40e.

The insidiousness of this problem is that everything works with the wrong
size, it's just not working as well as it could, as some MTU sizes end up
using two or more descriptors, and there is no way to tell that is
happening without looking at ice_trace or a bus analyzer.

Fixes: efc2214b6047 ("ice: Add support for XDP")
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Tested-by: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@intel.com> (A Contingent worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ipv4: fix data-races around inet->inet_id

[ Upstream commit f866fbc842de5976e41ba874b76ce31710b634b5 ]

UDP sendmsg() is lockless, so ip_select_ident_segs()
can very well be run from multiple cpus [1]

Convert inet->inet_id to an atomic_t, but implement
a dedicated path for TCP, avoiding cost of a locked
instruction (atomic_add_return())

Note that this patch will cause a trivial merge conflict
because we added inet->flags in net-next tree.

v2: added missing change in
drivers/net/ethernet/chelsio/inline_crypto/chtls/chtls_cm.c
(David Ahern)

[1]

BUG: KCSAN: data-race in __ip_make_skb / __ip_make_skb

read-write to 0xffff888145af952a of 2 bytes by task 7803 on cpu 1:
ip_select_ident_segs include/net/ip.h:542 [inline]
ip_select_ident include/net/ip.h:556 [inline]
__ip_make_skb+0x844/0xc70 net/ipv4/ip_output.c:1446
ip_make_skb+0x233/0x2c0 net/ipv4/ip_output.c:1560
udp_sendmsg+0x1199/0x1250 net/ipv4/udp.c:1260
inet_sendmsg+0x63/0x80 net/ipv4/af_inet.c:830
sock_sendmsg_nosec net/socket.c:725 [inline]
sock_sendmsg net/socket.c:748 [inline]
____sys_sendmsg+0x37c/0x4d0 net/socket.c:2494
___sys_sendmsg net/socket.c:2548 [inline]
__sys_sendmmsg+0x269/0x500 net/socket.c:2634
__do_sys_sendmmsg net/socket.c:2663 [inline]
__se_sys_sendmmsg net/socket.c:2660 [inline]
__x64_sys_sendmmsg+0x57/0x60 net/socket.c:2660
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x41/0xc0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd

read to 0xffff888145af952a of 2 bytes by task 7804 on cpu 0:
ip_select_ident_segs include/net/ip.h:541 [inline]
ip_select_ident include/net/ip.h:556 [inline]
__ip_make_skb+0x817/0xc70 net/ipv4/ip_output.c:1446
ip_make_skb+0x233/0x2c0 net/ipv4/ip_output.c:1560
udp_sendmsg+0x1199/0x1250 net/ipv4/udp.c:1260
inet_sendmsg+0x63/0x80 net/ipv4/af_inet.c:830
sock_sendmsg_nosec net/socket.c:725 [inline]
sock_sendmsg net/socket.c:748 [inline]
____sys_sendmsg+0x37c/0x4d0 net/socket.c:2494
___sys_sendmsg net/socket.c:2548 [inline]
__sys_sendmmsg+0x269/0x500 net/socket.c:2634
__do_sys_sendmmsg net/socket.c:2663 [inline]
__se_sys_sendmmsg net/socket.c:2660 [inline]
__x64_sys_sendmmsg+0x57/0x60 net/socket.c:2660
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x41/0xc0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd

value changed: 0x184d -> 0x184e

Reported by Kernel Concurrency Sanitizer on:
CPU: 0 PID: 7804 Comm: syz-executor.1 Not tainted 6.5.0-rc6-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/26/2023
==================================================================

Fixes: 23f57406b82d ("ipv4: avoid using shared IP generator for connected sockets")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: validate veth and vxcan peer ifindexes

[ Upstream commit f534f6581ec084fe94d6759f7672bd009794b07e ]

veth and vxcan need to make sure the ifindexes of the peer
are not negative, core does not validate this.

Using iproute2 with user-space-level checking removed:

Before:

  # ./ip link add index 10 type veth peer index -1
  # ip link show
  1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN mode DEFAULT group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
  2: enp1s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP mode DEFAULT group default qlen 1000
    link/ether 52:54:00:74:b2:03 brd ff:ff:ff:ff:ff:ff
  10: veth1@veth0: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
    link/ether 8a:90:ff:57:6d:5d brd ff:ff:ff:ff:ff:ff
  -1: veth0@veth1: <BROADCAST,MULTICAST,M-DOWN> mtu 1500 qdisc noop state DOWN mode DEFAULT group default qlen 1000
    link/ether ae:ed:18:e6:fa:7f brd ff:ff:ff:ff:ff:ff

Now:

  $ ./ip link add index 10 type veth peer index -1
  Error: ifindex can't be negative.

This problem surfaced in net-next because an explicit WARN()
was added, the root cause is older.

Fixes: e6f8f1a739b6 ("veth: Allow to create peer link with given ifindex")
Fixes: a8f820a380a2 ("can: add Virtual CAN Tunnel driver (vxcan)")
Reported-by: syzbot+5ba06978f34abb058571@syzkaller.appspotmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: bcmgenet: Fix return value check for fixed_phy_register()

[ Upstream commit 32bbe64a1386065ab2aef8ce8cae7c689d0add6e ]

The fixed_phy_register() function returns error pointers and never
returns NULL. Update the checks accordingly.

Fixes: b0ba512e25d7 ("net: bcmgenet: enable driver to work without a device tree")
Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Acked-by: Doug Berger <opendmb@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: bgmac: Fix return value check for fixed_phy_register()

[ Upstream commit 23a14488ea5882dea5851b65c9fce2127ee8fcad ]

The fixed_phy_register() function returns error pointers and never
returns NULL. Update the checks accordingly.

Fixes: c25b23b8a387 ("bgmac: register fixed PHY for ARM BCM470X / BCM5301X chipsets")
Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: mt7530: fix handling of 802.1X PAE frames

[ Upstream commit e94b590abfff2cdbf0bdaa7d9904364c8d480af5 ]

802.1X PAE frames are link-local frames, therefore they must be trapped to
the CPU port. Currently, the MT753X switches treat 802.1X PAE frames as
regular multicast frames, therefore flooding them to user ports. To fix
this, set 802.1X PAE frames to be trapped to the CPU port(s).

Fixes: b8f126a8d543 ("net-next: dsa: add dsa support for Mediatek MT7530 switch")
Signed-off-by: Arınç ÜNAL <arinc.unal@arinc9.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

selftests: mlxsw: Fix test failure on Spectrum-4

[ Upstream commit f520489e99a35b0a5257667274fbe9afd2d8c50b ]

Remove assumptions about shared buffer cell size and instead query the
cell size from devlink. Adjust the test to send small packets that fit
inside a single cell.

Tested on Spectrum-{1,2,3,4}.

Fixes: 4735402173e6 ("mlxsw: spectrum: Extend to support Spectrum-4 ASIC")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/f7dfbf3c4d1cb23838d9eb99bab09afaa320c4ca.1692268427.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mlxsw: Fix the size of 'VIRT_ROUTER_MSB'

[ Upstream commit 348c976be0a599918b88729def198a843701c9fe ]

The field 'virtual router' was extended to 12 bits in Spectrum-4.
Therefore, the element 'MLXSW_AFK_ELEMENT_VIRT_ROUTER_MSB' needs 3 bits for
Spectrum < 4 and 4 bits for Spectrum >= 4.

The elements are stored in an internal storage scratchpad. Currently, the
MSB is defined there as 3 bits. It means that for Spectrum-4, only 2K VRFs
can be used for multicast routing, as the highest bit is not really used by
the driver. Fix the definition of 'VIRT_ROUTER_MSB' to use 4 bits. Adjust
the definitions of 'virtual router' field in the blocks accordingly - use
'_avoid_size_check' for Spectrum-2 instead of for Spectrum-4. Fix the mask
in parse function to use 4 bits.

Fixes: 6d5d8ebb881c ("mlxsw: Rename virtual router flex key element")
Signed-off-by: Amit Cohen <amcohen@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/79bed2b70f6b9ed58d4df02e9798a23da648015b.1692268427.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mlxsw: reg: Fix SSPR register layout

[ Upstream commit 0dc63b9cfd4c5666ced52c829fdd65dcaeb9f0f1 ]

The two most significant bits of the "local_port" field in the SSPR
register are always cleared since they are overwritten by the deprecated
and overlapping "sub_port" field.

On systems with more than 255 local ports (e.g., Spectrum-4), this
results in the firmware maintaining invalid mappings between system port
and local port. Specifically, two different systems ports (0x1 and
0x101) point to the same local port (0x1), which eventually leads to
firmware errors.

Fix by removing the deprecated "sub_port" field.

Fixes: fd24b29a1b74 ("mlxsw: reg: Align existing registers to use extended local_port field")
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/9b909a3033c8d3d6f67f237306bef4411c5e6ae4.1692268427.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mlxsw: pci: Set time stamp fields also when its type is MIRROR_UTC

[ Upstream commit bc2de151ab6ad0762a04563527ec42e54dde572a ]

Currently, in Spectrum-2 and above, time stamps are extracted from the CQE
into the time stamp fields in 'struct mlxsw_skb_cb', only when the CQE
time stamp type is UTC. The time stamps are read directly from the CQE and
software can get the time stamp in UTC format using CQEv2.

From Spectrum-4, the time stamps that are read from the CQE are allowed
to be also from MIRROR_UTC type.

Therefore, we get a warning [1] from the driver that the time stamp fields
were not set, when LLDP control packet is sent.

Allow the time stamp type to be MIRROR_UTC and set the time stamp in this
case as well.

[1]
 WARNING: CPU: 11 PID: 0 at drivers/net/ethernet/mellanox/mlxsw/spectrum_ptp.c:1409 mlxsw_sp2_ptp_hwtstamp_fill+0x1f/0x70 [mlxsw_spectrum]
[...]
 Call Trace:
  <IRQ>
  mlxsw_sp2_ptp_receive+0x3c/0x80 [mlxsw_spectrum]
  mlxsw_core_skb_receive+0x119/0x190 [mlxsw_core]
  mlxsw_pci_cq_tasklet+0x3c9/0x780 [mlxsw_pci]
  tasklet_action_common.constprop.0+0x9f/0x110
  __do_softirq+0xbb/0x296
  irq_exit_rcu+0x79/0xa0
  common_interrupt+0x86/0xa0
  </IRQ>
  <TASK>

Fixes: 4735402173e6 ("mlxsw: spectrum: Extend to support Spectrum-4 ASIC")
Signed-off-by: Danielle Ratson <danieller@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/bcef4d044ef608a4e258d33a7ec0ecd91f480db5.1692268427.git.petrm@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ipvlan: Fix a reference count leak warning in ipvlan_ns_exit()

[ Upstream commit 043d5f68d0ccdda91029b4b6dce7eeffdcfad281 ]

There are two network devices(veth1 and veth3) in ns1, and ipvlan1 with
L3S mode and ipvlan2 with L2 mode are created based on them as
figure (1). In this case, ipvlan_register_nf_hook() will be called to
register nf hook which is needed by ipvlans in L3S mode in ns1 and value
of ipvl_nf_hook_refcnt is set to 1.

(1)
           ns1                           ns2
      ------------                  ------------

   veth1--ipvlan1 (L3S)

   veth3--ipvlan2 (L2)

(2)
           ns1                           ns2
      ------------                  ------------

   veth1--ipvlan1 (L3S)

         ipvlan2 (L2)                  veth3
     |                                  |
     |------->-------->--------->--------
                    migrate

When veth3 migrates from ns1 to ns2 as figure (2), veth3 will register in
ns2 and calls call_netdevice_notifiers with NETDEV_REGISTER event:

dev_change_net_namespace
    call_netdevice_notifiers
        ipvlan_device_event
            ipvlan_migrate_l3s_hook
                ipvlan_register_nf_hook(newnet)      (I)
                ipvlan_unregister_nf_hook(oldnet)    (II)

In function ipvlan_migrate_l3s_hook(), ipvl_nf_hook_refcnt in ns1 is not 0
since veth1 with ipvlan1 still in ns1, (I) and (II) will be called to
register nf_hook in ns2 and unregister nf_hook in ns1. As a result,
ipvl_nf_hook_refcnt in ns1 is decreased incorrectly and this in ns2
is increased incorrectly. When the second net namespace is removed, a
reference count leak warning in ipvlan_ns_exit() will be triggered.

This patch add a check before ipvlan_migrate_l3s_hook() is called. The
warning can be triggered as follows:

$ ip netns add ns1
$ ip netns add ns2
$ ip netns exec ns1 ip link add veth1 type veth peer name veth2
$ ip netns exec ns1 ip link add veth3 type veth peer name veth4
$ ip netns exec ns1 ip link add ipv1 link veth1 type ipvlan mode l3s
$ ip netns exec ns1 ip link add ipv2 link veth3 type ipvlan mode l2
$ ip netns exec ns1 ip link set veth3 netns ns2
$ ip net del ns2

Fixes: 3133822f5ac1 ("ipvlan: use pernet operations and restrict l3s hooks to master netns")
Signed-off-by: Lu Wei <luwei32@huawei.com>
Reviewed-by: Florian Westphal <fw@strlen.de>
Link: https://lore.kernel.org/r/20230817145449.141827-1-luwei32@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dccp: annotate data-races in dccp_poll()

[ Upstream commit cba3f1786916063261e3e5ccbb803abc325b24ef ]

We changed tcp_poll() over time, bug never updated dccp.

Note that we also could remove dccp instead of maintaining it.

Fixes: 7c657876b63c ("[DCCP]: Initial implementation")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://lore.kernel.org/r/20230818015820.2701595-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sock: annotate data-races around prot->memory_pressure

[ Upstream commit 76f33296d2e09f63118db78125c95ef56df438e9 ]

*prot->memory_pressure is read/writen locklessly, we need
to add proper annotations.

A recent commit added a new race, it is time to audit all accesses.

Fixes: 2d0c88e84e48 ("sock: Fix misuse of sk_under_memory_pressure()")
Fixes: 4d93df0abd50 ("[SCTP]: Rewrite of sctp buffer management code")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Abel Wu <wuyun.abel@bytedance.com>
Reviewed-by: Shakeel Butt <shakeelb@google.com>
Link: https://lore.kernel.org/r/20230818015132.2699348-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: felix: fix oversize frame dropping for always closed tc-taprio gates

[ Upstream commit d44036cad31170da0cb9c728e80743f84267da6e ]

The blamed commit resolved a bug where frames would still get stuck at
egress, even though they're smaller than the maxSDU[tc], because the
driver did not take into account the extra 33 ns that the queue system
needs for scheduling the frame.

It now takes that into account, but the arithmetic that we perform in
vsc9959_tas_remaining_gate_len_ps() is buggy, because we operate on
64-bit unsigned integers, so gate_len_ns - VSC9959_TAS_MIN_GATE_LEN_NS
may become a very large integer if gate_len_ns < 33 ns.

In practice, this means that we've introduced a regression where all
traffic class gates which are permanently closed will not get detected
by the driver, and we won't enable oversize frame dropping for them.

Before:
mscc_felix 0000:00:00.5: port 0: max frame size 1526 needs 12400000 ps, 1152000 ps for mPackets at speed 1000
mscc_felix 0000:00:00.5: port 0 tc 0 min gate len 1000000, sending all frames
mscc_felix 0000:00:00.5: port 0 tc 1 min gate len 0, sending all frames
mscc_felix 0000:00:00.5: port 0 tc 2 min gate len 0, sending all frames
mscc_felix 0000:00:00.5: port 0 tc 3 min gate len 0, sending all frames
mscc_felix 0000:00:00.5: port 0 tc 4 min gate len 0, sending all frames
mscc_felix 0000:00:00.5: port 0 tc 5 min gate len 0, sending all frames
mscc_felix 0000:00:00.5: port 0 tc 6 min gate len 0, sending all frames
mscc_felix 0000:00:00.5: port 0 tc 7 min gate length 5120 ns not enough for max frame size 1526 at 1000 Mbps, dropping frames over 615 octets including FCS

After:
mscc_felix 0000:00:00.5: port 0: max frame size 1526 needs 12400000 ps, 1152000 ps for mPackets at speed 1000
mscc_felix 0000:00:00.5: port 0 tc 0 min gate len 1000000, sending all frames
mscc_felix 0000:00:00.5: port 0 tc 1 min gate length 0 ns not enough for max frame size 1526 at 1000 Mbps, dropping frames over 1 octets including FCS
mscc_felix 0000:00:00.5: port 0 tc 2 min gate length 0 ns not enough for max frame size 1526 at 1000 Mbps, dropping frames over 1 octets including FCS
mscc_felix 0000:00:00.5: port 0 tc 3 min gate length 0 ns not enough for max frame size 1526 at 1000 Mbps, dropping frames over 1 octets including FCS
mscc_felix 0000:00:00.5: port 0 tc 4 min gate length 0 ns not enough for max frame size 1526 at 1000 Mbps, dropping frames over 1 octets including FCS
mscc_felix 0000:00:00.5: port 0 tc 5 min gate length 0 ns not enough for max frame size 1526 at 1000 Mbps, dropping frames over 1 octets including FCS
mscc_felix 0000:00:00.5: port 0 tc 6 min gate length 0 ns not enough for max frame size 1526 at 1000 Mbps, dropping frames over 1 octets including FCS
mscc_felix 0000:00:00.5: port 0 tc 7 min gate length 5120 ns not enough for max frame size 1526 at 1000 Mbps, dropping frames over 615 octets including FCS

Fixes: 11afdc6526de ("net: dsa: felix: tc-taprio intervals smaller than MTU should send at least one packet")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20230817120111.3522827-1-vladimir.oltean@nxp.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

devlink: add missing unregister linecard notification

[ Upstream commit 2ebbc9752d06bb1d01201fe632cb6da033b0248d ]

Cited fixes commit introduced linecard notifications for register,
however it didn't add them for unregister. Fix that by adding them.

Fixes: c246f9b5fd61 ("devlink: add support to create line card and expose to user")
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://lore.kernel.org/r/20230817125240.2144794-1-jiri@resnulli.us
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

devlink: move code to a dedicated directory

[ Upstream commit f05bd8ebeb69c803efd6d8a76d96b7fcd7011094 ]

The devlink code is hard to navigate with 13kLoC in one file.
I really like the way Michal split the ethtool into per-command
files and core. It'd probably be too much to split it all up,
but we can at least separate the core parts out of the per-cmd
implementations and put it in a directory so that new commands
can be separate files.

Move the code, subsequent commit will do a partial split.

Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stable-dep-of: 2ebbc9752d06 ("devlink: add missing unregister linecard notification")
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-af: SDP: fix receive link config

[ Upstream commit 05f3d5bc23524bed6f043dfe6b44da687584f9fb ]

On SDP interfaces, frame oversize and undersize errors are
observed as driver is not considering packet sizes of all
subscribers of the link before updating the link config.

This patch fixes the same.

Fixes: 9b7dd87ac071 ("octeontx2-af: Support to modify min/max allowed packet lengths")
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-by: Leon Romanovsky <leonro@nvidia.com>
Link: https://lore.kernel.org/r/20230817063006.10366-1-hkelam@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tracing: Fix memleak due to race between current_tracer and trace

[ Upstream commit eecb91b9f98d6427d4af5fdb8f108f52572a39e7 ]

Kmemleak report a leak in graph_trace_open():

  unreferenced object 0xffff0040b95f4a00 (size 128):
    comm "cat", pid 204981, jiffies 4301155872 (age 99771.964s)
    hex dump (first 32 bytes):
      e0 05 e7 b4 ab 7d 00 00 0b 00 01 00 00 00 00 00 .....}..........
      f4 00 01 10 00 a0 ff ff 00 00 00 00 65 00 10 00 ............e...
    backtrace:
      [<000000005db27c8b>] kmem_cache_alloc_trace+0x348/0x5f0
      [<000000007df90faa>] graph_trace_open+0xb0/0x344
      [<00000000737524cd>] __tracing_open+0x450/0xb10
      [<0000000098043327>] tracing_open+0x1a0/0x2a0
      [<00000000291c3876>] do_dentry_open+0x3c0/0xdc0
      [<000000004015bcd6>] vfs_open+0x98/0xd0
      [<000000002b5f60c9>] do_open+0x520/0x8d0
      [<00000000376c7820>] path_openat+0x1c0/0x3e0
      [<00000000336a54b5>] do_filp_open+0x14c/0x324
      [<000000002802df13>] do_sys_openat2+0x2c4/0x530
      [<0000000094eea458>] __arm64_sys_openat+0x130/0x1c4
      [<00000000a71d7881>] el0_svc_common.constprop.0+0xfc/0x394
      [<00000000313647bf>] do_el0_svc+0xac/0xec
      [<000000002ef1c651>] el0_svc+0x20/0x30
      [<000000002fd4692a>] el0_sync_handler+0xb0/0xb4
      [<000000000c309c35>] el0_sync+0x160/0x180

The root cause is descripted as follows:

  __tracing_open() {  // 1. File 'trace' is being opened;
    ...
    *iter->trace = *tr->current_trace;  // 2. Tracer 'function_graph' is
                                        //    currently set;
    ...
    iter->trace->open(iter);  // 3. Call graph_trace_open() here,
                              //    and memory are allocated in it;
    ...
  }

  s_start() {  // 4. The opened file is being read;
    ...
    *iter->trace = *tr->current_trace;  // 5. If tracer is switched to
                                        //    'nop' or others, then memory
                                        //    in step 3 are leaked!!!
    ...
  }

To fix it, in s_start(), close tracer before switching then reopen the
new tracer after switching. And some tracers like 'wakeup' may not update
'iter->private' in some cases when reopen, then it should be cleared
to avoid being mistakenly closed again.

Link: https://lore.kernel.org/linux-trace-kernel/20230817125539.1646321-1-zhengyejian1@huawei.com
Fixes: d7350c3f4569 ("tracing/core: make the read callbacks reentrants")
Signed-off-by: Zheng Yejian <zhengyejian1@huawei.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

tracing: Fix cpu buffers unavailable due to 'record_disabled' missed

[ Upstream commit b71645d6af10196c46cbe3732de2ea7d36b3ff6d ]

Trace ring buffer can no longer record anything after executing
following commands at the shell prompt:

  # cd /sys/kernel/tracing
  # cat tracing_cpumask
  fff
  # echo 0 > tracing_cpumask
  # echo 1 > snapshot
  # echo fff > tracing_cpumask
  # echo 1 > tracing_on
  # echo "hello world" > trace_marker
  -bash: echo: write error: Bad file descriptor

The root cause is that:
  1. After `echo 0 > tracing_cpumask`, 'record_disabled' of cpu buffers
     in 'tr->array_buffer.buffer' became 1 (see tracing_set_cpumask());
  2. After `echo 1 > snapshot`, 'tr->array_buffer.buffer' is swapped
     with 'tr->max_buffer.buffer', then the 'record_disabled' became 0
     (see update_max_tr());
  3. After `echo fff > tracing_cpumask`, the 'record_disabled' become -1;
Then array_buffer and max_buffer are both unavailable due to value of
'record_disabled' is not 0.

To fix it, enable or disable both array_buffer and max_buffer at the same
time in tracing_set_cpumask().

Link: https://lkml.kernel.org/r/20230805033816.3284594-2-zhengyejian1@huawei.com
Cc: <mhiramat@kernel.org>
Cc: <vnagarnaik@google.com>
Cc: <shuah@kernel.org>
Fixes: 71babb2705e2 ("tracing: change CPU ring buffer state from tracing_cpumask")
Signed-off-by: Zheng Yejian <zhengyejian1@huawei.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/i915/gt: Support aux invalidation on all engines

[ Upstream commit 6a35f22d222528e1b157c6978c9424d2f8cbe0a1 ]

Perform some refactoring with the purpose of keeping in one
single place all the operations around the aux table
invalidation.

With this refactoring add more engines where the invalidation
should be performed.

Fixes: 972282c4cf24 ("drm/i915/gen12: Add aux table invalidate for all engines")
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Cc: Jonathan Cavitt <jonathan.cavitt@intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: <stable@vger.kernel.org> # v5.8+
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230725001950.1014671-8-andi.shyti@linux.intel.com
(cherry picked from commit 76ff7789d6e63d1a10b3b58f5c70b2e640c7a880)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/i915/gt: Poll aux invalidation register bit on invalidation

[ Upstream commit 0fde2f23516a00fd90dfb980b66b4665fcbfa659 ]

For platforms that use Aux CCS, wait for aux invalidation to
complete by checking the aux invalidation register bit is
cleared.

Fixes: 972282c4cf24 ("drm/i915/gen12: Add aux table invalidate for all engines")
Signed-off-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Cc: <stable@vger.kernel.org> # v5.8+
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230725001950.1014671-7-andi.shyti@linux.intel.com
(cherry picked from commit d459c86f00aa98028d155a012c65dc42f7c37e76)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/i915/gt: Ensure memory quiesced before invalidation

[ Upstream commit 78a6ccd65fa3a7cc697810db079cc4b84dff03d5 ]

All memory traffic must be quiesced before requesting
an aux invalidation on platforms that use Aux CCS.

Fixes: 972282c4cf24 ("drm/i915/gen12: Add aux table invalidate for all engines")
Requires: a2a4aa0eef3b ("drm/i915: Add the gen12_needs_ccs_aux_inv helper")
Signed-off-by: Jonathan Cavitt <jonathan.cavitt@intel.com>
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Cc: <stable@vger.kernel.org> # v5.8+
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230725001950.1014671-4-andi.shyti@linux.intel.com
(cherry picked from commit ad8ebf12217e451cd19804b1c3e97ad56491c74a)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/i915: Add the gen12_needs_ccs_aux_inv helper

[ Upstream commit b2f59e9026038a5bbcbc0019fa58f963138211ee ]

We always assumed that a device might either have AUX or FLAT
CCS, but this is an approximation that is not always true, e.g.
PVC represents an exception.

Set the basis for future finer selection by implementing a
boolean gen12_needs_ccs_aux_inv() function that tells whether aux
invalidation is needed or not.

Currently PVC is the only exception to the above mentioned rule.

Requires: 059ae7ae2a1c ("drm/i915/gt: Cleanup aux invalidation registers")
Signed-off-by: Andi Shyti <andi.shyti@linux.intel.com>
Cc: Matt Roper <matthew.d.roper@intel.com>
Cc: Jonathan Cavitt <jonathan.cavitt@intel.com>
Cc: <stable@vger.kernel.org> # v5.8+
Reviewed-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Andrzej Hajda <andrzej.hajda@intel.com>
Reviewed-by: Nirmoy Das <nirmoy.das@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230725001950.1014671-3-andi.shyti@linux.intel.com
(cherry picked from commit c827655b87ad201ebe36f2e28d16b5491c8f7801)
Signed-off-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

s390/zcrypt: fix reply buffer calculations for CCA replies

[ Upstream commit 4cfca532ddc3474b3fc42592d0e4237544344b1a ]

The length information for available buffer space for CCA
replies is covered with two fields in the T6 header prepended
on each CCA reply: fromcardlen1 and fromcardlen2. The sum of
these both values must not exceed the AP bus limit for this
card (24KB for CEX8, 12KB CEX7 and older) minus the always
present headers.

The current code adjusted the fromcardlen2 value in case
of exceeding the AP bus limit when there was a non-zero
value given from userspace. Some tests now showed that this
was the wrong assumption. Instead the userspace value given for
this field should always be trusted and if the sum of the
two fields exceeds the AP bus limit for this card the first
field fromcardlen1 should be adjusted instead.

So now the calculation is done with this new insight in mind.
Also some additional checks for overflow have been introduced
and some comments to provide some documentation for future
maintainers of this complicated calculation code.

Furthermore the 128 bytes of fix overhead which is used
in the current code is not correct. Investigations showed
that for a reply always the same two header structs are
prepended before a possible payload. So this is also fixed
with this patch.

Signed-off-by: Harald Freudenberger <freude@linux.ibm.com>
Reviewed-by: Holger Dengler <dengler@linux.ibm.com>
Cc: stable@vger.kernel.org
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

s390/zcrypt: remove unnecessary (void *) conversions

[ Upstream commit 72c2112ce9d72e6c40dd893f32187a3d34453113 ]

Pointer variables of void * type do not require type cast.

Signed-off-by: Yu Zhe <yuzhe@nfschina.com>
Reviewed-by: Muhammad Usama Anjum <usama.anjum@collabora.com>
Link: https://lore.kernel.org/r/20230303052155.21072-1-yuzhe@nfschina.com
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Vasily Gorbik <gor@linux.ibm.com>
Stable-dep-of: 4cfca532ddc3 ("s390/zcrypt: fix reply buffer calculations for CCA replies")
Signed-off-by: Sasha Levin <sashal@kernel.org>

can: raw: fix lockdep issue in raw_release()

[ Upstream commit 11c9027c983e9e4b408ee5613b6504d24ebd85be ]

syzbot complained about a lockdep issue [1]

Since raw_bind() and raw_setsockopt() first get RTNL
before locking the socket, we must adopt the same order in raw_release()

[1]
WARNING: possible circular locking dependency detected
6.5.0-rc1-syzkaller-00192-g78adb4bcf99e #0 Not tainted
------------------------------------------------------
syz-executor.0/14110 is trying to acquire lock:
ffff88804e4b6130 (sk_lock-AF_CAN){+.+.}-{0:0}, at: lock_sock include/net/sock.h:1708 [inline]
ffff88804e4b6130 (sk_lock-AF_CAN){+.+.}-{0:0}, at: raw_bind+0xb1/0xab0 net/can/raw.c:435

but task is already holding lock:
ffffffff8e3df368 (rtnl_mutex){+.+.}-{3:3}, at: raw_bind+0xa7/0xab0 net/can/raw.c:434

which lock already depends on the new lock.

the existing dependency chain (in reverse order) is:

-> #1 (rtnl_mutex){+.+.}-{3:3}:
__mutex_lock_common kernel/locking/mutex.c:603 [inline]
__mutex_lock+0x181/0x1340 kernel/locking/mutex.c:747
raw_release+0x1c6/0x9b0 net/can/raw.c:391
__sock_release+0xcd/0x290 net/socket.c:654
sock_close+0x1c/0x20 net/socket.c:1386
__fput+0x3fd/0xac0 fs/file_table.c:384
task_work_run+0x14d/0x240 kernel/task_work.c:179
resume_user_mode_work include/linux/resume_user_mode.h:49 [inline]
exit_to_user_mode_loop kernel/entry/common.c:171 [inline]
exit_to_user_mode_prepare+0x210/0x240 kernel/entry/common.c:204
__syscall_exit_to_user_mode_work kernel/entry/common.c:286 [inline]
syscall_exit_to_user_mode+0x1d/0x50 kernel/entry/common.c:297
do_syscall_64+0x44/0xb0 arch/x86/entry/common.c:86
entry_SYSCALL_64_after_hwframe+0x63/0xcd

-> #0 (sk_lock-AF_CAN){+.+.}-{0:0}:
check_prev_add kernel/locking/lockdep.c:3142 [inline]
check_prevs_add kernel/locking/lockdep.c:3261 [inline]
validate_chain kernel/locking/lockdep.c:3876 [inline]
__lock_acquire+0x2e3d/0x5de0 kernel/locking/lockdep.c:5144
lock_acquire kernel/locking/lockdep.c:5761 [inline]
lock_acquire+0x1ae/0x510 kernel/locking/lockdep.c:5726
lock_sock_nested+0x3a/0xf0 net/core/sock.c:3492
lock_sock include/net/sock.h:1708 [inline]
raw_bind+0xb1/0xab0 net/can/raw.c:435
__sys_bind+0x1ec/0x220 net/socket.c:1792
__do_sys_bind net/socket.c:1803 [inline]
__se_sys_bind net/socket.c:1801 [inline]
__x64_sys_bind+0x72/0xb0 net/socket.c:1801
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd

other info that might help us debug this:

Possible unsafe locking scenario:

CPU0 CPU1
---- ----
lock(rtnl_mutex);
        lock(sk_lock-AF_CAN);
        lock(rtnl_mutex);
lock(sk_lock-AF_CAN);

*** DEADLOCK ***

1 lock held by syz-executor.0/14110:

stack backtrace:
CPU: 0 PID: 14110 Comm: syz-executor.0 Not tainted 6.5.0-rc1-syzkaller-00192-g78adb4bcf99e #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/03/2023
Call Trace:
<TASK>
__dump_stack lib/dump_stack.c:88 [inline]
dump_stack_lvl+0xd9/0x1b0 lib/dump_stack.c:106
check_noncircular+0x311/0x3f0 kernel/locking/lockdep.c:2195
check_prev_add kernel/locking/lockdep.c:3142 [inline]
check_prevs_add kernel/locking/lockdep.c:3261 [inline]
validate_chain kernel/locking/lockdep.c:3876 [inline]
__lock_acquire+0x2e3d/0x5de0 kernel/locking/lockdep.c:5144
lock_acquire kernel/locking/lockdep.c:5761 [inline]
lock_acquire+0x1ae/0x510 kernel/locking/lockdep.c:5726
lock_sock_nested+0x3a/0xf0 net/core/sock.c:3492
lock_sock include/net/sock.h:1708 [inline]
raw_bind+0xb1/0xab0 net/can/raw.c:435
__sys_bind+0x1ec/0x220 net/socket.c:1792
__do_sys_bind net/socket.c:1803 [inline]
__se_sys_bind net/socket.c:1801 [inline]
__x64_sys_bind+0x72/0xb0 net/socket.c:1801
do_syscall_x64 arch/x86/entry/common.c:50 [inline]
do_syscall_64+0x38/0xb0 arch/x86/entry/common.c:80
entry_SYSCALL_64_after_hwframe+0x63/0xcd
RIP: 0033:0x7fd89007cb29
Code: 28 00 00 00 75 05 48 83 c4 28 c3 e8 e1 20 00 00 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 73 01 c3 48 c7 c1 b0 ff ff ff f7 d8 64 89 01 48
RSP: 002b:00007fd890d2a0c8 EFLAGS: 00000246 ORIG_RAX: 0000000000000031
RAX: ffffffffffffffda RBX: 00007fd89019bf80 RCX: 00007fd89007cb29
RDX: 0000000000000010 RSI: 0000000020000040 RDI: 0000000000000003
RBP: 00007fd8900c847a R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 0000000000000000
R13: 000000000000000b R14: 00007fd89019bf80 R15: 00007ffebf8124f8
</TASK>

Fixes: ee8b94c8510c ("can: raw: fix receiver memory leak")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ziyang Xuan <william.xuanziyang@huawei.com>
Cc: Oliver Hartkopp <socketcan@hartkopp.net>
Cc: stable@vger.kernel.org
Cc: Marc Kleine-Budde <mkl@pengutronix.de>
Link: https://lore.kernel.org/all/20230720114438.172434-1-edumazet@google.com
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

can: raw: fix receiver memory leak

[ Upstream commit ee8b94c8510ce64afe0b87ef548d23e00915fb10 ]

Got kmemleak errors with the following ltp can_filter testcase:

for ((i=1; i<=100; i++))
do
        ./can_filter &
        sleep 0.1
done

==============================================================
[<00000000db4a4943>] can_rx_register+0x147/0x360 [can]
[<00000000a289549d>] raw_setsockopt+0x5ef/0x853 [can_raw]
[<000000006d3d9ebd>] __sys_setsockopt+0x173/0x2c0
[<00000000407dbfec>] __x64_sys_setsockopt+0x61/0x70
[<00000000fd468496>] do_syscall_64+0x33/0x40
[<00000000b7e47d51>] entry_SYSCALL_64_after_hwframe+0x61/0xc6

It's a bug in the concurrent scenario of unregister_netdevice_many()
and raw_release() as following:

             cpu0                                        cpu1
unregister_netdevice_many(can_dev)
  unlist_netdevice(can_dev) // dev_get_by_index() return NULL after this
  net_set_todo(can_dev)
raw_release(can_socket)
  dev = dev_get_by_index(, ro->ifindex); // dev == NULL
  if (dev) { // receivers in dev_rcv_lists not free because dev is NULL
    raw_disable_allfilters(, dev, );
    dev_put(dev);
  }
  ...
  ro->bound = 0;
  ...

call_netdevice_notifiers(NETDEV_UNREGISTER, )
  raw_notify(, NETDEV_UNREGISTER, )
    if (ro->bound) // invalid because ro->bound has been set 0
      raw_disable_allfilters(, dev, ); // receivers in dev_rcv_lists will never be freed

Add a net_device pointer member in struct raw_sock to record bound
can_dev, and use rtnl_lock to serialize raw_socket members between
raw_bind(), raw_release(), raw_setsockopt() and raw_notify(). Use
ro->dev to decide whether to free receivers in dev_rcv_lists.

Fixes: 8d0caedb7596 ("can: bcm/raw/isotp: use per module netdevice notifier")
Reviewed-by: Oliver Hartkopp <socketcan@hartkopp.net>
Acked-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Ziyang Xuan <william.xuanziyang@huawei.com>
Link: https://lore.kernel.org/all/20230711011737.1969582-1-william.xuanziyang@huawei.com
Cc: stable@vger.kernel.org
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

jbd2: fix a race when checking checkpoint buffer busy

[ Upstream commit 46f881b5b1758dc4a35fba4a643c10717d0cf427 ]

Before removing checkpoint buffer from the t_checkpoint_list, we have to
check both BH_Dirty and BH_Lock bits together to distinguish buffers
have not been or were being written back. But __cp_buffer_busy() checks
them separately, it first check lock state and then check dirty, the
window between these two checks could be raced by writing back
procedure, which locks buffer and clears buffer dirty before I/O
completes. So it cannot guarantee checkpointing buffers been written
back to disk if some error happens later. Finally, it may clean
checkpoint transactions and lead to inconsistent filesystem.

jbd2_journal_forget() and __journal_try_to_free_buffer() also have the
same problem (journal_unmap_buffer() escape from this issue since it's
running under the buffer lock), so fix them through introducing a new
helper to try holding the buffer lock and remove really clean buffer.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=217490
Cc: stable@vger.kernel.org
Suggested-by: Jan Kara <jack@suse.cz>
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230606135928.434610-6-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Signed-off-by: Sasha Levin <sashal@kernel.org>

jbd2: remove journal_clean_one_cp_list()

[ Upstream commit b98dba273a0e47dbfade89c9af73c5b012a4eabb ]

journal_clean_one_cp_list() and journal_shrink_one_cp_list() are almost
the same, so merge them into journal_shrink_one_cp_list(), remove the
nr_to_scan parameter, always scan and try to free the whole checkpoint
list.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230606135928.434610-4-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Stable-dep-of: 46f881b5b175 ("jbd2: fix a race when checking checkpoint buffer busy")
Signed-off-by: Sasha Levin <sashal@kernel.org>

jbd2: remove t_checkpoint_io_list

[ Upstream commit be22255360f80d3af789daad00025171a65424a5 ]

Since t_checkpoint_io_list was stop using in jbd2_log_do_checkpoint()
now, it's time to remove the whole t_checkpoint_io_list logic.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230606135928.434610-3-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Stable-dep-of: 46f881b5b175 ("jbd2: fix a race when checking checkpoint buffer busy")
Signed-off-by: Sasha Levin <sashal@kernel.org>

MIPS: cpu-features: Use boot_cpu_type for CPU type based features

[ Upstream commit 5487a7b60695a92cf998350e4beac17144c91fcd ]

Some CPU feature macros were using current_cpu_type to mark feature
availability.

However current_cpu_type will use smp_processor_id, which is prohibited
under preemptable context.

Since those features are all uniform on all CPUs in a SMP system, use
boot_cpu_type instead of current_cpu_type to fix preemptable kernel.

Cc: stable@vger.kernel.org
Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

MIPS: cpu-features: Enable octeon_cache by cpu_type

[ Upstream commit f641519409a73403ee6612b8648b95a688ab85c2 ]

cpu_has_octeon_cache was tied to 0 for generic cpu-features,
whith this generic kernel built for octeon CPU won't boot.

Just enable this flag by cpu_type. It won't hurt orther platforms
because compiler will eliminate the code path on other processors.

Signed-off-by: Jiaxun Yang <jiaxun.yang@flygoat.com>
Signed-off-by: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Stable-dep-of: 5487a7b60695 ("MIPS: cpu-features: Use boot_cpu_type for CPU type based features")
Signed-off-by: Sasha Levin <sashal@kernel.org>

PCI: acpiphp: Reassign resources on bridge if necessary

[ Upstream commit 40613da52b13fb21c5566f10b287e0ca8c12c4e9 ]

When using ACPI PCI hotplug, hotplugging a device with large BARs may fail
if bridge windows programmed by firmware are not large enough.

Reproducer:
  $ qemu-kvm -monitor stdio -M q35  -m 4G \
      -global ICH9-LPC.acpi-pci-hotplug-with-bridge-support=on \
      -device id=rp1,pcie-root-port,bus=pcie.0,chassis=4 \
      disk_image

 wait till linux guest boots, then hotplug device:
   (qemu) device_add qxl,bus=rp1

 hotplug on guest side fails with:
   pci 0000:01:00.0: [1b36:0100] type 00 class 0x038000
   pci 0000:01:00.0: reg 0x10: [mem 0x00000000-0x03ffffff]
   pci 0000:01:00.0: reg 0x14: [mem 0x00000000-0x03ffffff]
   pci 0000:01:00.0: reg 0x18: [mem 0x00000000-0x00001fff]
   pci 0000:01:00.0: reg 0x1c: [io  0x0000-0x001f]
   pci 0000:01:00.0: BAR 0: no space for [mem size 0x04000000]
   pci 0000:01:00.0: BAR 0: failed to assign [mem size 0x04000000]
   pci 0000:01:00.0: BAR 1: no space for [mem size 0x04000000]
   pci 0000:01:00.0: BAR 1: failed to assign [mem size 0x04000000]
   pci 0000:01:00.0: BAR 2: assigned [mem 0xfe800000-0xfe801fff]
   pci 0000:01:00.0: BAR 3: assigned [io  0x1000-0x101f]
   qxl 0000:01:00.0: enabling device (0000 -> 0003)
   Unable to create vram_mapping
   qxl: probe of 0000:01:00.0 failed with error -12

However when using native PCIe hotplug
  '-global ICH9-LPC.acpi-pci-hotplug-with-bridge-support=off'
it works fine, since kernel attempts to reassign unused resources.

Use the same machinery as native PCIe hotplug to (re)assign resources.

Link: https://lore.kernel.org/r/20230424191557.2464760-1-imammedo@redhat.com
Signed-off-by: Igor Mammedov <imammedo@redhat.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Acked-by: Rafael J. Wysocki <rafael@kernel.org>
Cc: stable@vger.kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

video/aperture: Move vga handling to pci function

[ Upstream commit f1d599d315fb7b7343cddaf365e671aaa8453aca ]

A few reasons for this:

- It's really the only one where this matters. I tried looking around,
  and I didn't find any non-pci vga-compatible controllers for x86
  (since that's the only platform where we had this until a few
  patches ago), where a driver participating in the aperture claim
  dance would interfere.

- I also don't expect that any future bus anytime soon will
  not just look like pci towards the OS, that's been the case for like
  25+ years by now for practically everything (even non non-x86).

- Also it's a bit funny if we have one part of the vga removal in the
  pci function, and the other in the generic one.

v2: Rebase.

v4:
- fix Daniel's S-o-b address

v5:
- add back an S-o-b tag with Daniel's Intel address

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Javier Martinez Canillas <javierm@redhat.com>
Cc: Helge Deller <deller@gmx.de>
Cc: linux-fbdev@vger.kernel.org
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230406132109.32050-6-tzimmermann@suse.de
Stable-dep-of: 5ae3716cfdcd ("video/aperture: Only remove sysfb on the default vga pci device")
Signed-off-by: Sasha Levin <sashal@kernel.org>

video/aperture: Only kick vgacon when the pdev is decoding vga

[ Upstream commit 7450cd235b45d43ee6f3c235f89e92623458175d ]

Otherwise it's a bit silly, and we might throw out the driver for the
screen the user is actually looking at. I haven't found a bug report
for this case yet, but we did get bug reports for the analog case
where we're throwing out the efifb driver.

v2: Flip the check around to make it clear it's a special case for
kicking out the vgacon driver only (Thomas)

v4:
- fixes to commit message
- fix Daniel's S-o-b address

v5:
- add back an S-o-b tag with Daniel's Intel address

Link: https://bugzilla.kernel.org/show_bug.cgi?id=216303
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Javier Martinez Canillas <javierm@redhat.com>
Cc: Helge Deller <deller@gmx.de>
Cc: linux-fbdev@vger.kernel.org
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230406132109.32050-5-tzimmermann@suse.de
Stable-dep-of: 5ae3716cfdcd ("video/aperture: Only remove sysfb on the default vga pci device")
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/aperture: Remove primary argument

[ Upstream commit 62aeaeaa1b267c5149abee6b45967a5df3feed58 ]

Only really pci devices have a business setting this - it's for
figuring out whether the legacy vga stuff should be nuked too. And
with the preceding two patches those are all using the pci version of
this.

Which means for all other callers primary == false and we can remove
it now.

v2:
- Reorder to avoid compile fail (Thomas)
- Include gma500, which retained it's called to the non-pci version.

v4:
- fix Daniel's S-o-b address

v5:
- add back an S-o-b tag with Daniel's Intel address

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Javier Martinez Canillas <javierm@redhat.com>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Maxime Ripard <mripard@kernel.org>
Cc: Deepak Rawat <drawat.floss@gmail.com>
Cc: Neil Armstrong <neil.armstrong@linaro.org>
Cc: Kevin Hilman <khilman@baylibre.com>
Cc: Jerome Brunet <jbrunet@baylibre.com>
Cc: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Cc: Thierry Reding <thierry.reding@gmail.com>
Cc: Jonathan Hunter <jonathanh@nvidia.com>
Cc: Emma Anholt <emma@anholt.net>
Cc: Helge Deller <deller@gmx.de>
Cc: David Airlie <airlied@gmail.com>
Cc: Daniel Vetter <daniel@ffwll.ch>
Cc: linux-hyperv@vger.kernel.org
Cc: linux-amlogic@lists.infradead.org
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-tegra@vger.kernel.org
Cc: linux-fbdev@vger.kernel.org
Acked-by: Martin Blumenstingl <martin.blumenstingl@googlemail.com>
Acked-by: Thierry Reding <treding@nvidia.com>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230406132109.32050-4-tzimmermann@suse.de
Stable-dep-of: 5ae3716cfdcd ("video/aperture: Only remove sysfb on the default vga pci device")
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/gma500: Use drm_aperture_remove_conflicting_pci_framebuffers

[ Upstream commit 80e993988b97fe794f3ec2be6db05fe30f9353c3 ]

This one nukes all framebuffers, which is a bit much. In reality
gma500 is igpu and never shipped with anything discrete, so there should
not be any difference.

v2: Unfortunately the framebuffer sits outside of the pci bars for
gma500, and so only using the pci helpers won't be enough. Otoh if we
only use non-pci helper, then we don't get the vga handling, and
subsequent refactoring to untangle these special cases won't work.

It's not pretty, but the simplest fix (since gma500 really is the only
quirky pci driver like this we have) is to just have both calls.

v4:
- fix Daniel's S-o-b address

v5:
- add back an S-o-b tag with Daniel's Intel address

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Javier Martinez Canillas <javierm@redhat.com>
Reviewed-by: Javier Martinez Canillas <javierm@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230406132109.32050-2-tzimmermann@suse.de
Stable-dep-of: 5ae3716cfdcd ("video/aperture: Only remove sysfb on the default vga pci device")
Signed-off-by: Sasha Levin <sashal@kernel.org>

fbdev/radeon: use pci aperture helpers

[ Upstream commit 9b539c4d1b921bc9c8c578d4d50f0a7e7874d384 ]

It's not exactly the same since the open coded version doesn't set
primary correctly. But that's a bugfix, so shouldn't hurt really.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org>
Cc: linux-fbdev@vger.kernel.org
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20230111154112.90575-7-daniel.vetter@ffwll.ch
Stable-dep-of: 5ae3716cfdcd ("video/aperture: Only remove sysfb on the default vga pci device")
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/ast: Use drm_aperture_remove_conflicting_pci_framebuffers

[ Upstream commit c1ebead36099deb85384f6fb262fe619a04cee73 ]

It's just open coded and matches.

Note that Thomas said that his version apparently failed for some
reason, but hey maybe we should try again.

Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Cc: Dave Airlie <airlied@redhat.com>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: Javier Martinez Canillas <javierm@redhat.com>
Cc: Helge Deller <deller@gmx.de>
Cc: linux-fbdev@vger.kernel.org
Tested-by: Thomas Zimmmermann <tzimmermann@suse.de>
Reviewed-by: Thomas Zimmermann <tzimmermann@suse.de>
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Link: https://patchwork.freedesktop.org/patch/msgid/20230111154112.90575-1-daniel.vetter@ffwll.ch
Stable-dep-of: 5ae3716cfdcd ("video/aperture: Only remove sysfb on the default vga pci device")
Signed-off-by: Sasha Levin <sashal@kernel.org>

xprtrdma: Remap Receive buffers after a reconnect

[ Upstream commit 895cedc1791916e8a98864f12b656702fad0bb67 ]

On server-initiated disconnect, rpcrdma_xprt_disconnect() was DMA-
unmapping the Receive buffers, but rpcrdma_post_recvs() neglected
to remap them after a new connection had been established. The
result was immediate failure of the new connection with the Receives
flushing with LOCAL_PROT_ERR.

Fixes: 671c450b6fe0 ("xprtrdma: Fix oops in Receive handler after device removal")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

NFSv4: fix out path in __nfs4_get_acl_uncached

[ Upstream commit f4e89f1a6dab4c063fc1e823cc9dddc408ff40cf ]

Another highly rare error case when a page allocating loop (inside
__nfs4_get_acl_uncached, this time) is not properly unwound on error.
Since pages array is allocated being uninitialized, need to free only
lower array indices. NULL checks were useful before commit 62a1573fcf84
("NFSv4 fix acl retrieval over krb5i/krb5p mounts") when the array had
been initialized to zero on stack.

Found by Linux Verification Center (linuxtesting.org).

Fixes: 62a1573fcf84 ("NFSv4 fix acl retrieval over krb5i/krb5p mounts")
Signed-off-by: Fedor Pchelkin <pchelkin@ispras.ru>
Reviewed-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

NFSv4.2: fix error handling in nfs42_proc_getxattr

[ Upstream commit 4e3733fd2b0f677faae21cf838a43faf317986d3 ]

There is a slight issue with error handling code inside
nfs42_proc_getxattr(). If page allocating loop fails then we free the
failing page array element which is NULL but __free_page() can't deal with
NULL args.

Found by Linux Verification Center (linuxtesting.org).

Fixes: a1f26739ccdc ("NFSv4.2: improve page handling for GETXATTR")
Signed-off-by: Fedor Pchelkin <pchelkin@ispras.ru>
Reviewed-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Linux 6.1.49

Link: https://lore.kernel.org/r/20230826154625.450325166@linuxfoundation.org
Tested-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Tested-by: Takeshi Ogasawara <takeshi.ogasawara@futuring-girl.com>
Tested-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Salvatore Bonaccorso <carnil@debian.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "f2fs: fix to do sanity check on direct node in truncate_dnode()"

This reverts commit a78a8bcdc26de5ef3a0ee27c9c6c512e54a6051c which is
commit a6ec83786ab9f13f25fb18166dee908845713a95 upstream.

Something is currently broken in the f2fs code, Guenter has reported
boot problems with it for a few releases now, so revert the most recent
f2fs changes in the hope to get this back to a working filesystem.

Reported-by: Guenter Roeck <linux@roeck-us.net>
Link: https://lore.kernel.org/r/b392e1a8-b987-4993-bd45-035db9415a6e@roeck-us.net
Cc: Chao Yu <chao@kernel.org>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "f2fs: fix to set flush_merge opt and show noflush_merge"

This reverts commit 6ba0594a81f91d6fd8ca9bd4ad23aa1618635a0f which is
commit 967eaad1fed5f6335ea97a47d45214744dc57925 upstream.

Something is currently broken in the f2fs code, Guenter has reported
boot problems with it for a few releases now, so revert the most recent
f2fs changes in the hope to get this back to a working filesystem.

Reported-by: Guenter Roeck <linux@roeck-us.net>
Link: https://lore.kernel.org/r/b392e1a8-b987-4993-bd45-035db9415a6e@roeck-us.net
Cc: Chao Yu <chao@kernel.org>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Yangtao Li <frank.li@vivo.com>
Cc: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "f2fs: don't reset unchangable mount option in f2fs_remount()"

This reverts commit e2fb24ce37caeaecff08af4e9967c8462624312b which is
commit 458c15dfbce62c35fefd9ca637b20a051309c9f1 upstream.

Something is currently broken in the f2fs code, Guenter has reported
boot problems with it for a few releases now, so revert the most recent
f2fs changes in the hope to get this back to a working filesystem.

Reported-by: Guenter Roeck <linux@roeck-us.net>
Link: https://lore.kernel.org/r/b392e1a8-b987-4993-bd45-035db9415a6e@roeck-us.net
Cc: Chao Yu <chao@kernel.org>
Cc: Jaegeuk Kim <jaegeuk@kernel.org>
Cc: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

objtool/x86: Fix SRSO mess

commit 4ae68b26c3ab5a82aa271e6e9fc9b1a06e1d6b40 upstream.

Objtool --rethunk does two things:

 - it collects all (tail) call's of __x86_return_thunk and places them
   into .return_sites. These are typically compiler generated, but
   RET also emits this same.

 - it fudges the validation of the __x86_return_thunk symbol; because
   this symbol is inside another instruction, it can't actually find
   the instruction pointed to by the symbol offset and gets upset.

Because these two things pertained to the same symbol, there was no
pressing need to separate these two separate things.

However, alas, along comes SRSO and more crazy things to deal with
appeared.

The SRSO patch itself added the following symbol names to identify as
rethunk:

  'srso_untrain_ret', 'srso_safe_ret' and '__ret'

Where '__ret' is the old retbleed return thunk, 'srso_safe_ret' is a
new similarly embedded return thunk, and 'srso_untrain_ret' is
completely unrelated to anything the above does (and was only included
because of that INT3 vs UD2 issue fixed previous).

Clear things up by adding a second category for the embedded instruction
thing.

Fixes: fb3bd914b3ec ("x86/srso: Add a Speculative RAS Overflow mitigation")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230814121148.704502245@infradead.org
Signed-off-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 6.1.48

Link: https://lore.kernel.org/r/20230824141447.155846739@linuxfoundation.org
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: SeongJae Park <sj@kernel.org>
Tested-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Tested-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Conor Dooley <conor.dooley@microchip.com>
Tested-by: Takeshi Ogasawara <takeshi.ogasawara@futuring-girl.com>
Tested-by: Shuah Khan <skhan@linuxfoundation.org>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Tested-by: Salvatore Bonaccorso <carnil@debian.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/srso: Correct the mitigation status when SMT is disabled

commit 6405b72e8d17bd1875a56ae52d23ec3cd51b9d66 upstream.

Specify how is SRSO mitigated when SMT is disabled. Also, correct the
SMT check for that.

Fixes: e9fbc47b818b ("x86/srso: Disable the mitigation on unaffected configurations")
Suggested-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Acked-by: Josh Poimboeuf <jpoimboe@kernel.org>
Link: https://lore.kernel.org/r/20230814200813.p5czl47zssuej7nv@treble
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

objtool/x86: Fixup frame-pointer vs rethunk

commit dbf46008775516f7f25c95b7760041c286299783 upstream.

For stack-validation of a frame-pointer build, objtool validates that
every CALL instruction is preceded by a frame-setup. The new SRSO
return thunks violate this with their RSB stuffing trickery.

Extend the __fentry__ exception to also cover the embedded_insn case
used for this. This cures:

  vmlinux.o: warning: objtool: srso_untrain_ret+0xd: call without frame pointer save/setup

Fixes: 4ae68b26c3ab ("objtool/x86: Fix SRSO mess")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Acked-by: Josh Poimboeuf <jpoimboe@kernel.org>
Link: https://lore.kernel.org/r/20230816115921.GH980931@hirez.programming.kicks-ass.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/retpoline,kprobes: Fix position of thunk sections with CONFIG_LTO_CLANG

commit 79cd2a11224eab86d6673fe8a11d2046ae9d2757 upstream.

The linker script arch/x86/kernel/vmlinux.lds.S matches the thunk
sections ".text.__x86.*" from arch/x86/lib/retpoline.S as follows:

  .text {
    [...]
    TEXT_TEXT
    [...]
    __indirect_thunk_start = .;
    *(.text.__x86.*)
    __indirect_thunk_end = .;
    [...]
  }

Macro TEXT_TEXT references TEXT_MAIN which normally expands to only
".text". However, with CONFIG_LTO_CLANG, TEXT_MAIN becomes
".text .text.[0-9a-zA-Z_]*" which wrongly matches also the thunk
sections. The output layout is then different than expected. For
instance, the currently defined range [__indirect_thunk_start,
__indirect_thunk_end] becomes empty.

Prevent the problem by using ".." as the first separator, for example,
".text..__x86.indirect_thunk". This pattern is utilized by other
explicit section names which start with one of the standard prefixes,
such as ".text" or ".data", and that need to be individually selected in
the linker script.

  [ nathan: Fix conflicts with SRSO and fold in fix issue brought up by
    Andrew Cooper in post-review:
    https://lore.kernel.org/20230803230323.1478869-1-andrew.cooper3@citrix.com ]

Fixes: dc5723b02e52 ("kbuild: add support for Clang LTO")
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230711091952.27944-2-petr.pavlu@suse.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/srso: Disable the mitigation on unaffected configurations

commit e9fbc47b818b964ddff5df5b2d5c0f5f32f4a147 upstream.

Skip the srso cmd line parsing which is not needed on Zen1/2 with SMT
disabled and with the proper microcode applied (latter should be the
case anyway) as those are not affected.

Fixes: 5a15d8348881 ("x86/srso: Tie SBPB bit setting to microcode patch detection")
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230813104517.3346-1-bp@alien8.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/CPU/AMD: Fix the DIV(0) initial fix attempt

commit f58d6fbcb7c848b7f2469be339bc571f2e9d245b upstream.

Initially, it was thought that doing an innocuous division in the #DE
handler would take care to prevent any leaking of old data from the
divider but by the time the fault is raised, the speculation has already
advanced too far and such data could already have been used by younger
operations.

Therefore, do the innocuous division on every exit to userspace so that
userspace doesn't see any potentially old data from integer divisions in
kernel space.

Do the same before VMRUN too, to protect host data from leaking into the
guest too.

Fixes: 77245f1c3c64 ("x86/CPU/AMD: Do not leak quotient data after a division by 0")
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Cc: <stable@kernel.org>
Link: https://lore.kernel.org/r/20230811213824.10025-1-bp@alien8.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/retpoline: Don't clobber RFLAGS during srso_safe_ret()

commit ba5ca5e5e6a1d55923e88b4a83da452166f5560e upstream.

Use LEA instead of ADD when adjusting %rsp in srso_safe_ret{,_alias}()
so as to avoid clobbering flags.  Drop one of the INT3 instructions to
account for the LEA consuming one more byte than the ADD.

KVM's emulator makes indirect calls into a jump table of sorts, where
the destination of each call is a small blob of code that performs fast
emulation by executing the target instruction with fixed operands.

E.g. to emulate ADC, fastop() invokes adcb_al_dl():

  adcb_al_dl:
    <+0>:  adc    %dl,%al
    <+2>:  jmp    <__x86_return_thunk>

A major motivation for doing fast emulation is to leverage the CPU to
handle consumption and manipulation of arithmetic flags, i.e. RFLAGS is
both an input and output to the target of the call.  fastop() collects
the RFLAGS result by pushing RFLAGS onto the stack and popping them back
into a variable (held in %rdi in this case):

  asm("push %[flags]; popf; " CALL_NOSPEC " ; pushf; pop %[flags]\n"

  <+71>: mov    0xc0(%r8),%rdx
  <+78>: mov    0x100(%r8),%rcx
  <+85>: push   %rdi
  <+86>: popf
  <+87>: call   *%rsi
  <+89>: nop
  <+90>: nop
  <+91>: nop
  <+92>: pushf
  <+93>: pop    %rdi

and then propagating the arithmetic flags into the vCPU's emulator state:

  ctxt->eflags = (ctxt->eflags & ~EFLAGS_MASK) | (flags & EFLAGS_MASK);

  <+64>:  and    $0xfffffffffffff72a,%r9
  <+94>:  and    $0x8d5,%edi
  <+109>: or     %rdi,%r9
  <+122>: mov    %r9,0x10(%r8)

The failures can be most easily reproduced by running the "emulator"
test in KVM-Unit-Tests.

If you're feeling a bit of deja vu, see commit b63f20a778c8
("x86/retpoline: Don't clobber RFLAGS during CALL_NOSPEC on i386").

In addition, this breaks booting of clang-compiled guest on
a gcc-compiled host where the host contains the %rsp-modifying SRSO
mitigations.

  [ bp: Massage commit message, extend, remove addresses. ]

Fixes: fb3bd914b3ec ("x86/srso: Add a Speculative RAS Overflow mitigation")
Closes: https://lore.kernel.org/all/de474347-122d-54cd-eabf-9dcc95ab9eae@amd.com
Reported-by: Srikanth Aithal <sraithal@amd.com>
Reported-by: Nathan Chancellor <nathan@kernel.org>
Signed-off-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Tested-by: Nathan Chancellor <nathan@kernel.org>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/20230810013334.GA5354@dev-arch.thelio-3990X/
Link: https://lore.kernel.org/r/20230811155255.250835-1-seanjc@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/static_call: Fix __static_call_fixup()

commit 54097309620ef0dc2d7083783dc521c6a5fef957 upstream.

Christian reported spurious module load crashes after some of Song's
module memory layout patches.

Turns out that if the very last instruction on the very last page of the
module is a 'JMP __x86_return_thunk' then __static_call_fixup() will
trip a fault and die.

And while the module rework made this slightly more likely to happen,
it's always been possible.

Fixes: ee88d363d156 ("x86,static_call: Use alternative RET encoding")
Reported-by: Christian Bricart <christian@bricart.de>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Josh Poimboeuf <jpoimboe@kernel.org>
Link: https://lkml.kernel.org/r/20230816104419.GA982867@hirez.programming.kicks-ass.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/srso: Explain the untraining sequences a bit more

commit 9dbd23e42ff0b10c9b02c9e649c76e5228241a8e upstream.

The goal is to eventually have a proper documentation about all this.

Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230814164447.GFZNpZ/64H4lENIe94@fat_crate.local
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/cpu: Cleanup the untrain mess

commit e7c25c441e9e0fa75b4c83e0b26306b702cfe90d upstream.

Since there can only be one active return_thunk, there only needs be
one (matching) untrain_ret. It fundamentally doesn't make sense to
allow multiple untrain_ret at the same time.

Fold all the 3 different untrain methods into a single (temporary)
helper stub.

Fixes: fb3bd914b3ec ("x86/srso: Add a Speculative RAS Overflow mitigation")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230814121149.042774962@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/cpu: Rename srso_(.*)_alias to srso_alias_\1

commit 42be649dd1f2eee6b1fb185f1a231b9494cf095f upstream.

For a more consistent namespace.

  [ bp: Fixup names in the doc too. ]

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230814121148.976236447@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/cpu: Rename original retbleed methods

commit d025b7bac07a6e90b6b98b487f88854ad9247c39 upstream.

Rename the original retbleed return thunk and untrain_ret to
retbleed_return_thunk() and retbleed_untrain_ret().

No functional changes.

Suggested-by: Josh Poimboeuf <jpoimboe@kernel.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230814121148.909378169@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/cpu: Clean up SRSO return thunk mess

commit d43490d0ab824023e11d0b57d0aeec17a6e0ca13 upstream.

Use the existing configurable return thunk. There is absolute no
justification for having created this __x86_return_thunk alternative.

To clarify, the whole thing looks like:

Zen3/4 does:

  srso_alias_untrain_ret:
  nop2
  lfence
  jmp srso_alias_return_thunk
  int3

  srso_alias_safe_ret: // aliasses srso_alias_untrain_ret just so
  add $8, %rsp
  ret
  int3

  srso_alias_return_thunk:
  call srso_alias_safe_ret
  ud2

While Zen1/2 does:

  srso_untrain_ret:
  movabs $foo, %rax
  lfence
  call srso_safe_ret           (jmp srso_return_thunk ?)
  int3

  srso_safe_ret: // embedded in movabs instruction
  add $8,%rsp
          ret
          int3

  srso_return_thunk:
  call srso_safe_ret
  ud2

While retbleed does:

  zen_untrain_ret:
  test $0xcc, %bl
  lfence
  jmp zen_return_thunk
          int3

  zen_return_thunk: // embedded in the test instruction
  ret
          int3

Where Zen1/2 flush the BTB entry using the instruction decoder trick
(test,movabs) Zen3/4 use BTB aliasing. SRSO adds a return sequence
(srso_safe_ret()) which forces the function return instruction to
speculate into a trap (UD2).  This RET will then mispredict and
execution will continue at the return site read from the top of the
stack.

Pick one of three options at boot (evey function can only ever return
once).

  [ bp: Fixup commit message uarch details and add them in a comment in
    the code too. Add a comment about the srso_select_mitigation()
    dependency on retbleed_select_mitigation(). Add moar ifdeffery for
    32-bit builds. Add a dummy srso_untrain_ret_alias() definition for
    32-bit alternatives needing the symbol. ]

Fixes: fb3bd914b3ec ("x86/srso: Add a Speculative RAS Overflow mitigation")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230814121148.842775684@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/alternative: Make custom return thunk unconditional

commit 095b8303f3835c68ac4a8b6d754ca1c3b6230711 upstream.

There is infrastructure to rewrite return thunks to point to any
random thunk one desires, unwrap that from CALL_THUNKS, which up to
now was the sole user of that.

  [ bp: Make the thunks visible on 32-bit and add ifdeffery for the
    32-bit builds. ]

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230814121148.775293785@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/cpu: Fix up srso_safe_ret() and __x86_return_thunk()

commit af023ef335f13c8b579298fc432daeef609a9e60 upstream.

  vmlinux.o: warning: objtool: srso_untrain_ret() falls through to next function __x86_return_skl()
  vmlinux.o: warning: objtool: __x86_return_thunk() falls through to next function __x86_return_skl()

This is because these functions (can) end with CALL, which objtool
does not consider a terminating instruction. Therefore, replace the
INT3 instruction (which is a non-fatal trap) with UD2 (which is a
fatal-trap).

This indicates execution will not continue past this point.

Fixes: fb3bd914b3ec ("x86/srso: Add a Speculative RAS Overflow mitigation")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230814121148.637802730@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/cpu: Fix __x86_return_thunk symbol type

commit 77f67119004296a9b2503b377d610e08b08afc2a upstream.

Commit

  fb3bd914b3ec ("x86/srso: Add a Speculative RAS Overflow mitigation")

reimplemented __x86_return_thunk with a mix of SYM_FUNC_START and
SYM_CODE_END, this is not a sane combination.

Since nothing should ever actually 'CALL' this, make it consistently
CODE.

Fixes: fb3bd914b3ec ("x86/srso: Add a Speculative RAS Overflow mitigation")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Link: https://lore.kernel.org/r/20230814121148.571027074@infradead.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 6.1.47

Link: https://lore.kernel.org/r/20230821194122.695845670@linuxfoundation.org
Tested-by: Joel Fernandes (Google) <joel@joelfernandes.org>
Tested-by: Takeshi Ogasawara <takeshi.ogasawara@futuring-girl.com>
Tested-by: Bagas Sanjaya <bagasdotme@gmail.com>
Tested-by: SeongJae Park <sj@kernel.org>
Tested-by: Shuah Khan <skhan@linuxfoundation.org>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: Ron Economos <re@w6rz.net>
Tested-by: Sudip Mukherjee <sudip.mukherjee@codethink.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: f-sdh30: fix order of function calls in sdhci_f_sdh30_remove

commit 58abdd80b93b09023ca03007b608685c41e3a289 upstream.

The order of function calls in sdhci_f_sdh30_remove is wrong,
let's call sdhci_pltfm_unregister first.

Cc: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Fixes: 5def5c1c15bf ("mmc: sdhci-f-sdh30: Replace with sdhci_pltfm")
Signed-off-by: Yangtao Li <frank.li@vivo.com>
Reported-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Acked-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20230727070051.17778-62-frank.li@vivo.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: fix the RTO timer retransmitting skb every 1ms if linear option is enabled

commit e4dd0d3a2f64b8bd8029ec70f52bdbebd0644408 upstream.

In the real workload, I encountered an issue which could cause the RTO
timer to retransmit the skb per 1ms with linear option enabled. The amount
of lost-retransmitted skbs can go up to 1000+ instantly.

The root cause is that if the icsk_rto happens to be zero in the 6th round
(which is the TCP_THIN_LINEAR_RETRIES value), then it will always be zero
due to the changed calculation method in tcp_retransmit_timer() as follows:

icsk->icsk_rto = min(icsk->icsk_rto << 1, TCP_RTO_MAX);

Above line could be converted to
icsk->icsk_rto = min(0 << 1, TCP_RTO_MAX) = 0

Therefore, the timer expires so quickly without any doubt.

I read through the RFC 6298 and found that the RTO value can be rounded
up to a certain value, in Linux, say TCP_RTO_MIN as default, which is
regarded as the lower bound in this patch as suggested by Eric.

Fixes: 36e31b0af587 ("net: TCP thin linear timeouts")
Suggested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: Jason Xing <kernelxing@tencent.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/nouveau/disp: fix use-after-free in error handling of nouveau_connector_create

commit 1b254b791d7b7dea6e8adc887fbbd51746d8bb27 upstream.

We can't simply free the connector after calling drm_connector_init on it.
We need to clean up the drm side first.

It might not fix all regressions from commit 2b5d1c29f6c4
("drm/nouveau/disp: PIOR DP uses GPIO for HPD, not PMGR AUX interrupts"),
but at least it fixes a memory corruption in error handling related to
that commit.

Link: https://lore.kernel.org/lkml/20230806213107.GFZNARG6moWpFuSJ9W@fat_crate.local/
Fixes: 95983aea8003 ("drm/nouveau/disp: add connector class")
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20230814144933.3956959-1-kherbst@redhat.com
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

af_unix: Fix null-ptr-deref in unix_stream_sendpage().

Bing-Jhong Billy Jheng reported null-ptr-deref in unix_stream_sendpage()
with detailed analysis and a nice repro.

unix_stream_sendpage() tries to add data to the last skb in the peer's
recv queue without locking the queue.

If the peer's FD is passed to another socket and the socket's FD is
passed to the peer, there is a loop between them.  If we close both
sockets without receiving FD, the sockets will be cleaned up by garbage
collection.

The garbage collection iterates such sockets and unlinks skb with
FD from the socket's receive queue under the queue's lock.

So, there is a race where unix_stream_sendpage() could access an skb
locklessly that is being released by garbage collection, resulting in
use-after-free.

To avoid the issue, unix_stream_sendpage() must lock the peer's recv
queue.

Note the issue does not exist in 6.5+ thanks to the recent sendpage()
refactoring.

This patch is originally written by Linus Torvalds.

BUG: unable to handle page fault for address: ffff988004dd6870
PF: supervisor read access in kernel mode
PF: error_code(0x0000) - not-present page
PGD 0 P4D 0
PREEMPT SMP PTI
CPU: 4 PID: 297 Comm: garbage_uaf Not tainted 6.1.46 #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.16.0-0-gd239552ce722-prebuilt.qemu.org 04/01/2014
RIP: 0010:kmem_cache_alloc_node+0xa2/0x1e0
Code: c0 0f 84 32 01 00 00 41 83 fd ff 74 10 48 8b 00 48 c1 e8 3a 41 39 c5 0f 85 1c 01 00 00 41 8b 44 24 28 49 8b 3c 24 48 8d 4a 40 <49> 8b 1c 06 4c 89 f0 65 48 0f c7 0f 0f 94 c0 84 c0 74 a1 41 8b 44
RSP: 0018:ffffc9000079fac0 EFLAGS: 00000246
RAX: 0000000000000070 RBX: 0000000000000005 RCX: 000000000001a284
RDX: 000000000001a244 RSI: 0000000000400cc0 RDI: 000000000002eee0
RBP: 0000000000400cc0 R08: 0000000000400cc0 R09: 0000000000000003
R10: 0000000000000001 R11: 0000000000000000 R12: ffff888003970f00
R13: 00000000ffffffff R14: ffff988004dd6800 R15: 00000000000000e8
FS:  00007f174d6f3600(0000) GS:ffff88807db00000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: ffff988004dd6870 CR3: 00000000092be000 CR4: 00000000007506e0
PKRU: 55555554
Call Trace:
 <TASK>
 ? __die_body.cold+0x1a/0x1f
 ? page_fault_oops+0xa9/0x1e0
 ? fixup_exception+0x1d/0x310
 ? exc_page_fault+0xa8/0x150
 ? asm_exc_page_fault+0x22/0x30
 ? kmem_cache_alloc_node+0xa2/0x1e0
 ? __alloc_skb+0x16c/0x1e0
 __alloc_skb+0x16c/0x1e0
 alloc_skb_with_frags+0x48/0x1e0
 sock_alloc_send_pskb+0x234/0x270
 unix_stream_sendmsg+0x1f5/0x690
 sock_sendmsg+0x5d/0x60
 ____sys_sendmsg+0x210/0x260
 ___sys_sendmsg+0x83/0xd0
 ? kmem_cache_alloc+0xc6/0x1c0
 ? avc_disable+0x20/0x20
 ? percpu_counter_add_batch+0x53/0xc0
 ? alloc_empty_file+0x5d/0xb0
 ? alloc_file+0x91/0x170
 ? alloc_file_pseudo+0x94/0x100
 ? __fget_light+0x9f/0x120
 __sys_sendmsg+0x54/0xa0
 do_syscall_64+0x3b/0x90
 entry_SYSCALL_64_after_hwframe+0x69/0xd3
RIP: 0033:0x7f174d639a7d
Code: 28 89 54 24 1c 48 89 74 24 10 89 7c 24 08 e8 8a c1 f4 ff 8b 54 24 1c 48 8b 74 24 10 41 89 c0 8b 7c 24 08 b8 2e 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 33 44 89 c7 48 89 44 24 08 e8 de c1 f4 ff 48
RSP: 002b:00007ffcb563ea50 EFLAGS: 00000293 ORIG_RAX: 000000000000002e
RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007f174d639a7d
RDX: 0000000000000000 RSI: 00007ffcb563eab0 RDI: 0000000000000007
RBP: 00007ffcb563eb10 R08: 0000000000000000 R09: 00000000ffffffff
R10: 00000000004040a0 R11: 0000000000000293 R12: 00007ffcb563ec28
R13: 0000000000401398 R14: 0000000000403e00 R15: 00007f174d72c000
 </TASK>

Fixes: 869e7c62486e ("net: af_unix: implement stream sendpage support")
Reported-by: Bing-Jhong Billy Jheng <billy@starlabs.sg>
Reviewed-by: Bing-Jhong Billy Jheng <billy@starlabs.sg>
Co-developed-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amdgpu: keep irq count in amdgpu_irq_disable_all

commit 8ffd6f0442674f32c048ec8dffdbc5ec67829beb upstream.

This can clean up all irq warnings because of unbalanced
amdgpu_irq_get/put when unplugging/unbinding device, and leave
irq count decrease in each ip fini function.

Signed-off-by: Guchun Chen <guchun.chen@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drm/amd/pm: skip the RLC stop when S0i3 suspend for SMU v13.0.4/11

commit 730d44e1fa306a20746ad4a85da550662aed9daa upstream.

For SMU v13.0.4/11, driver does not need to stop RLC for S0i3,
the firmwares will handle that properly.

Signed-off-by: Tim Huang <Tim.Huang@amd.com>
Reviewed-by: Mario Limonciello <mario.limonciello@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

arm64/ptrace: Ensure that SME is set up for target when writing SSVE state

commit 5d0a8d2fba50e9c07cde4aad7fba28c008b07a5b upstream.

When we use NT_ARM_SSVE to either enable streaming mode or change the
vector length for a process we do not currently do anything to ensure that
there is storage allocated for the SME specific register state.  If the
task had not previously used SME or we changed the vector length then
the task will not have had TIF_SME set or backing storage for ZA/ZT
allocated, resulting in inconsistent register sizes when saving state
and spurious traps which flush the newly set register state.

We should set TIF_SME to disable traps and ensure that storage is
allocated for ZA and ZT if it is not already allocated.  This requires
modifying sme_alloc() to make the flush of any existing register state
optional so we don't disturb existing state for ZA and ZT.

Fixes: e12310a0d30f ("arm64/sme: Implement ptrace support for streaming mode SVE registers")
Reported-by: David Spickett <David.Spickett@arm.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Cc: <stable@vger.kernel.org> # 5.19.x
Link: https://lore.kernel.org/r/20230810-arm64-fix-ptrace-race-v1-1-a5361fad2bd6@kernel.org
Signed-off-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

netfilter: set default timeout to 3 secs for sctp shutdown send and recv state

commit 9bfab6d23a2865966a4f89a96536fbf23f83bc8c upstream.

In SCTP protocol, it is using the same timer (T2 timer) for SHUTDOWN and
SHUTDOWN_ACK retransmission. However in sctp conntrack the default timeout
value for SCTP_CONNTRACK_SHUTDOWN_ACK_SENT state is 3 secs while it's 300
msecs for SCTP_CONNTRACK_SHUTDOWN_SEND/RECV state.

As Paolo Valerio noticed, this might cause unwanted expiration of the ct
entry. In my test, with 1s tc netem delay set on the NAT path, after the
SHUTDOWN is sent, the sctp ct entry enters SCTP_CONNTRACK_SHUTDOWN_SEND
state. However, due to 300ms (too short) delay, when the SHUTDOWN_ACK is
sent back from the peer, the sctp ct entry has expired and been deleted,
and then the SHUTDOWN_ACK has to be dropped.

Also, it is confusing these two sysctl options always show 0 due to all
timeout values using sec as unit:

  net.netfilter.nf_conntrack_sctp_timeout_shutdown_recd = 0
  net.netfilter.nf_conntrack_sctp_timeout_shutdown_sent = 0

This patch fixes it by also using 3 secs for sctp shutdown send and recv
state in sctp conntrack, which is also RTO.initial value in SCTP protocol.

Note that the very short time value for SCTP_CONNTRACK_SHUTDOWN_SEND/RECV
was probably used for a rare scenario where SHUTDOWN is sent on 1st path
but SHUTDOWN_ACK is replied on 2nd path, then a new connection started
immediately on 1st path. So this patch also moves from SHUTDOWN_SEND/RECV
to CLOSE when receiving INIT in the ORIGINAL direction.

Fixes: 9fb9cbb1082d ("[NETFILTER]: Add nf_conntrack subsystem.")
Reported-by: Paolo Valerio <pvalerio@redhat.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

hugetlb: do not clear hugetlb dtor until allocating vmemmap

commit 32c877191e022b55fe3a374f3d7e9fb5741c514d upstream.

Patch series "Fix hugetlb free path race with memory errors".

In the discussion of Jiaqi Yan's series "Improve hugetlbfs read on
HWPOISON hugepages" the race window was discovered.
https://lore.kernel.org/linux-mm/20230616233447.GB7371@monkey/

Freeing a hugetlb page back to low level memory allocators is performed
in two steps.
1) Under hugetlb lock, remove page from hugetlb lists and clear destructor
2) Outside lock, allocate vmemmap if necessary and call low level free
Between these two steps, the hugetlb page will appear as a normal
compound page.  However, vmemmap for tail pages could be missing.
If a memory error occurs at this time, we could try to update page
flags non-existant page structs.

A much more detailed description is in the first patch.

The first patch addresses the race window.  However, it adds a
hugetlb_lock lock/unlock cycle to every vmemmap optimized hugetlb page
free operation.  This could lead to slowdowns if one is freeing a large
number of hugetlb pages.

The second path optimizes the update_and_free_pages_bulk routine to only
take the lock once in bulk operations.

The second patch is technically not a bug fix, but includes a Fixes tag
and Cc stable to avoid a performance regression.  It can be combined with
the first, but was done separately make reviewing easier.

This patch (of 2):

Freeing a hugetlb page and releasing base pages back to the underlying
allocator such as buddy or cma is performed in two steps:
- remove_hugetlb_folio() is called to remove the folio from hugetlb
  lists, get a ref on the page and remove hugetlb destructor.  This
  all must be done under the hugetlb lock.  After this call, the page
  can be treated as a normal compound page or a collection of base
  size pages.
- update_and_free_hugetlb_folio() is called to allocate vmemmap if
  needed and the free routine of the underlying allocator is called
  on the resulting page.  We can not hold the hugetlb lock here.

One issue with this scheme is that a memory error could occur between
these two steps.  In this case, the memory error handling code treats
the old hugetlb page as a normal compound page or collection of base
pages.  It will then try to SetPageHWPoison(page) on the page with an
error.  If the page with error is a tail page without vmemmap, a write
error will occur when trying to set the flag.

Address this issue by modifying remove_hugetlb_folio() and
update_and_free_hugetlb_folio() such that the hugetlb destructor is not
cleared until after allocating vmemmap.  Since clearing the destructor
requires holding the hugetlb lock, the clearing is done in
remove_hugetlb_folio() if the vmemmap is present.  This saves a
lock/unlock cycle.  Otherwise, destructor is cleared in
update_and_free_hugetlb_folio() after allocating vmemmap.

Note that this will leave hugetlb pages in a state where they are marked
free (by hugetlb specific page flag) and have a ref count.  This is not
a normal state.  The only code that would notice is the memory error
code, and it is set up to retry in such a case.

A subsequent patch will create a routine to do bulk processing of
vmemmap allocation.  This will eliminate a lock/unlock cycle for each
hugetlb page in the case where we are freeing a large number of pages.

Link: https://lkml.kernel.org/r/20230711220942.43706-1-mike.kravetz@oracle.com
Link: https://lkml.kernel.org/r/20230711220942.43706-2-mike.kravetz@oracle.com
Fixes: ad2fa3717b74 ("mm: hugetlb: alloc the vmemmap pages associated with each HugeTLB page")
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Reviewed-by: Muchun Song <songmuchun@bytedance.com>
Tested-by: Naoya Horiguchi <naoya.horiguchi@nec.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: James Houghton <jthoughton@google.com>
Cc: Jiaqi Yan <jiaqiyan@google.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>