| Age | Commit message (Collapse) | Author | Files | Lines |
|---|
|
Link: https://lore.kernel.org/r/20260408175913.177092714@linuxfoundation.org
Tested-by: Shung-Hsi Yu <shung-hsi.yu@suse.com>
Tested-by: Peter Schneider <pschneider1968@googlemail.com>
Tested-by: Francesco Dolcini <francesco.dolcini@toradex.com>
Tested-by: Ron Economos <re@w6rz.net>
Tested-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Mark Brown <broonie@kernel.org>
Tested-by: Miguel Ojeda <ojeda@kernel.org>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Tested-by: Shuah Khan <skhan@linuxfoundation.org>
Tested-by: Barry K. Nathan <barryn@pobox.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit eeee5a710f26ce57807024ef330fe5a850eaecd8 ]
In commit 8110633db49d7de2 ("net: sfp-bus: allow SFP quirks to override
Autoneg and pause bits") we moved the setting of Autoneg and pause bits
before the call to SFP quirk when parsing SFP module support.
Since the quirk for Ubiquiti U-Fiber Instant SFP module zeroes the
support bits and sets 1000baseX_Full only, the above mentioned commit
changed the overall computed support from
1000baseX_Full, Autoneg, Pause, Asym_Pause
to just
1000baseX_Full.
This broke the SFP module for mvneta, which requires Autoneg for
1000baseX since commit c762b7fac1b249a9 ("net: mvneta: deny disabling
autoneg for 802.3z modes").
Fix this by setting back the Autoneg, Pause and Asym_Pause bits in the
quirk.
Fixes: 8110633db49d7de2 ("net: sfp-bus: allow SFP quirks to override Autoneg and pause bits")
Signed-off-by: Marek Behún <kabel@kernel.org>
Reviewed-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Link: https://patch.msgid.link/20260326122038.2489589-1-kabel@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 7ab4a7c5d969642782b8a5b608da0dd02aa9f229 upstream.
In mptcp_pm_nl_create_listen_socket(), use entry->addr.family
instead of sk->sk_family for lock class setup. The 'sk' parameter
is a netlink socket, not the MPTCP subflow socket being created.
Fixes: cee4034a3db1 ("mptcp: fix lockdep false positive in mptcp_pm_nl_create_listen_socket()")
Signed-off-by: Li Xiasong <lixiasong1@huawei.com>
Reviewed-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Link: https://patch.msgid.link/20260319112159.3118874-1-lixiasong1@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ Conflict in pm_kernel.c, because commit 8617e85e04bd ("mptcp: pm:
split in-kernel PM specific code") is not in this version, and moves
code from pm_netlink.c to pm_kernel.c. ]
Signed-off-by: Matthieu Baerts (NGI0) <matttbe@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit bb81702370fad22c06ca12b6e1648754dbc37e0f ]
Commit 4865c768b563 ("ext4: always allocate blocks only from groups
inode can use") restricts what blocks will be allocated for indirect
block based files to block numbers that fit within 32-bit block
numbers.
However, when using a review bot running on the latest Gemini LLM to
check this commit when backporting into an LTS based kernel, it raised
this concern:
If ac->ac_g_ex.fe_group is >= ngroups (for instance, if the goal
group was populated via stream allocation from s_mb_last_groups),
then start will be >= ngroups.
Does this allow allocating blocks beyond the 32-bit limit for
indirect block mapped files? The commit message mentions that
ext4_mb_scan_groups_linear() takes care to not select unsupported
groups. However, its loop uses group = *start, and the very first
iteration will call ext4_mb_scan_group() with this unsupported
group because next_linear_group() is only called at the end of the
iteration.
After reviewing the code paths involved and considering the LLM
review, I determined that this can happen when there is a file system
where some files/directories are extent-mapped and others are
indirect-block mapped. To address this, add a safety clamp in
ext4_mb_scan_groups().
Fixes: 4865c768b563 ("ext4: always allocate blocks only from groups inode can use")
Cc: Jan Kara <jack@suse.cz>
Reviewed-by: Baokun Li <libaokun@linux.alibaba.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Link: https://patch.msgid.link/20260326045834.1175822-1-tytso@mit.edu
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 1aec30021edd410b986c156f195f3d23959a9d11 ]
ext4_inode_attach_jinode() publishes ei->jinode to concurrent users.
It used to set ei->jinode before jbd2_journal_init_jbd_inode(),
allowing a reader to observe a non-NULL jinode with i_vfs_inode
still unset.
The fast commit flush path can then pass this jinode to
jbd2_wait_inode_data(), which dereferences i_vfs_inode->i_mapping and
may crash.
Below is the crash I observe:
```
BUG: unable to handle page fault for address: 000000010beb47f4
PGD 110e51067 P4D 110e51067 PUD 0
Oops: Oops: 0000 [#1] SMP NOPTI
CPU: 1 UID: 0 PID: 4850 Comm: fc_fsync_bench_ Not tainted 6.18.0-00764-g795a690c06a5 #1 PREEMPT(voluntary)
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Arch Linux 1.17.0-2-2 04/01/2014
RIP: 0010:xas_find_marked+0x3d/0x2e0
Code: e0 03 48 83 f8 02 0f 84 f0 01 00 00 48 8b 47 08 48 89 c3 48 39 c6 0f 82 fd 01 00 00 48 85 c9 74 3d 48 83 f9 03 77 63 4c 8b 0f <49> 8b 71 08 48 c7 47 18 00 00 00 00 48 89 f1 83 e1 03 48 83 f9 02
RSP: 0018:ffffbbee806e7bf0 EFLAGS: 00010246
RAX: 000000000010beb4 RBX: 000000000010beb4 RCX: 0000000000000003
RDX: 0000000000000001 RSI: 0000002000300000 RDI: ffffbbee806e7c10
RBP: 0000000000000001 R08: 0000002000300000 R09: 000000010beb47ec
R10: ffff9ea494590090 R11: 0000000000000000 R12: 0000002000300000
R13: ffffbbee806e7c90 R14: ffff9ea494513788 R15: ffffbbee806e7c88
FS: 00007fc2f9e3e6c0(0000) GS:ffff9ea6b1444000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 000000010beb47f4 CR3: 0000000119ac5000 CR4: 0000000000750ef0
PKRU: 55555554
Call Trace:
<TASK>
filemap_get_folios_tag+0x87/0x2a0
__filemap_fdatawait_range+0x5f/0xd0
? srso_alias_return_thunk+0x5/0xfbef5
? __schedule+0x3e7/0x10c0
? srso_alias_return_thunk+0x5/0xfbef5
? srso_alias_return_thunk+0x5/0xfbef5
? srso_alias_return_thunk+0x5/0xfbef5
? preempt_count_sub+0x5f/0x80
? srso_alias_return_thunk+0x5/0xfbef5
? cap_safe_nice+0x37/0x70
? srso_alias_return_thunk+0x5/0xfbef5
? preempt_count_sub+0x5f/0x80
? srso_alias_return_thunk+0x5/0xfbef5
filemap_fdatawait_range_keep_errors+0x12/0x40
ext4_fc_commit+0x697/0x8b0
? ext4_file_write_iter+0x64b/0x950
? srso_alias_return_thunk+0x5/0xfbef5
? preempt_count_sub+0x5f/0x80
? srso_alias_return_thunk+0x5/0xfbef5
? vfs_write+0x356/0x480
? srso_alias_return_thunk+0x5/0xfbef5
? preempt_count_sub+0x5f/0x80
ext4_sync_file+0xf7/0x370
do_fsync+0x3b/0x80
? syscall_trace_enter+0x108/0x1d0
__x64_sys_fdatasync+0x16/0x20
do_syscall_64+0x62/0x2c0
entry_SYSCALL_64_after_hwframe+0x76/0x7e
...
```
Fix this by initializing the jbd2_inode first.
Use smp_wmb() and WRITE_ONCE() to publish ei->jinode after
initialization. Readers use READ_ONCE() to fetch the pointer.
Fixes: a361293f5fede ("jbd2: Fix oops in jbd2_journal_file_inode()")
Cc: stable@vger.kernel.org
Signed-off-by: Li Chen <me@linux.beauty>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://patch.msgid.link/20260225082617.147957-1-me@linux.beauty
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
Cc: stable@kernel.org
[ adapted READ_ONCE(ei->jinode) to use pos->jinode ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 2e7b5cf72e51c9cf9c8b75190189c757df31ddd9 ]
Configure only the requested channel when a fixed channel is specified
to avoid modifying other channels unintentionally.
Fix parameter configuration when a fixed DMA channel is requested on
i.MX9 AON domain and i.MX8QM/QXP/DXL platforms. When a client requests
a fixed channel (e.g., channel 6), the driver traverses channels 0-5
and may unintentionally modify their configuration if they are unused.
This leads to issues such as setting the `is_multi_fifo` flag unexpectedly,
causing memcpy tests to fail when using the dmatest tool.
Only affect edma memcpy test when the channel is fixed.
Fixes: 72f5801a4e2b ("dmaengine: fsl-edma: integrate v3 support")
Signed-off-by: Joy Zou <joy.zou@nxp.com>
Cc: stable@vger.kernel.org
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://patch.msgid.link/20250917-b4-edma-chanconf-v1-1-886486e02e91@nxp.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 90d21f6e57a898ef02810404dd3866acaf707ebf ]
Introduce a scope guard to automatically unlock the mutex within
fsl_edma3_xlate() to simplify the code.
Prepare to add source ID checks in the future.
Signed-off-by: Joy Zou <joy.zou@nxp.com>
Reviewed-by: Frank Li <Frank.Li@nxp.com>
Link: https://lore.kernel.org/r/20240701070232.2519179-2-joy.zou@nxp.com
Signed-off-by: Vinod Koul <vkoul@kernel.org>
Stable-dep-of: 2e7b5cf72e51 ("dmaengine: fsl-edma: fix channel parameter config for fixed channel requests")
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 05243d490bb7852a8acca7b5b5658019c7797a52 ]
Move FSGSBASE enablement from identify_cpu() to cpu_init_exception_handling()
to ensure it is enabled before any exceptions can occur on both boot and
secondary CPUs.
== Background ==
Exception entry code (paranoid_entry()) uses ALTERNATIVE patching based on
X86_FEATURE_FSGSBASE to decide whether to use RDGSBASE/WRGSBASE instructions
or the slower RDMSR/SWAPGS sequence for saving/restoring GSBASE.
On boot CPU, ALTERNATIVE patching happens after enabling FSGSBASE in CR4.
When the feature is available, the code is permanently patched to use
RDGSBASE/WRGSBASE, which require CR4.FSGSBASE=1 to execute without triggering
== Boot Sequence ==
Boot CPU (with CR pinning enabled):
trap_init()
cpu_init() <- Uses unpatched code (RDMSR/SWAPGS)
x2apic_setup()
...
arch_cpu_finalize_init()
identify_boot_cpu()
identify_cpu()
cr4_set_bits(X86_CR4_FSGSBASE) # Enables the feature
# This becomes part of cr4_pinned_bits
...
alternative_instructions() <- Patches code to use RDGSBASE/WRGSBASE
Secondary CPUs (with CR pinning enabled):
start_secondary()
cr4_init() <- Code already patched, CR4.FSGSBASE=1
set implicitly via cr4_pinned_bits
cpu_init() <- exceptions work because FSGSBASE is
already enabled
Secondary CPU (with CR pinning disabled):
start_secondary()
cr4_init() <- Code already patched, CR4.FSGSBASE=0
cpu_init()
x2apic_setup()
rdmsrq(MSR_IA32_APICBASE) <- Triggers #VC in SNP guests
exc_vmm_communication()
paranoid_entry() <- Uses RDGSBASE with CR4.FSGSBASE=0
(patched code)
...
ap_starting()
identify_secondary_cpu()
identify_cpu()
cr4_set_bits(X86_CR4_FSGSBASE) <- Enables the feature, which is
too late
== CR Pinning ==
Currently, for secondary CPUs, CR4.FSGSBASE is set implicitly through
CR-pinning: the boot CPU sets it during identify_cpu(), it becomes part of
cr4_pinned_bits, and cr4_init() applies those pinned bits to secondary CPUs.
This works but creates an undocumented dependency between cr4_init() and the
pinning mechanism.
== Problem ==
Secondary CPUs boot after alternatives have been applied globally. They
execute already-patched paranoid_entry() code that uses RDGSBASE/WRGSBASE
instructions, which require CR4.FSGSBASE=1. Upcoming changes to CR pinning
behavior will break the implicit dependency, causing secondary CPUs to
generate #UD.
This issue manifests itself on AMD SEV-SNP guests, where the rdmsrq() in
x2apic_setup() triggers a #VC exception early during cpu_init(). The #VC
handler (exc_vmm_communication()) executes the patched paranoid_entry() path.
Without CR4.FSGSBASE enabled, RDGSBASE instructions trigger #UD.
== Fix ==
Enable FSGSBASE explicitly in cpu_init_exception_handling() before loading
exception handlers. This makes the dependency explicit and ensures both
boot and secondary CPUs have FSGSBASE enabled before paranoid_entry()
executes.
Fixes: c82965f9e530 ("x86/entry/64: Handle FSGSBASE enabled paranoid entry/exit")
Reported-by: Borislav Petkov <bp@alien8.de>
Suggested-by: Sohil Mehta <sohil.mehta@intel.com>
Signed-off-by: Nikunj A Dadhania <nikunj@amd.com>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Sohil Mehta <sohil.mehta@intel.com>
Cc: <stable@kernel.org>
Link: https://patch.msgid.link/20260318075654.1792916-2-nikunj@amd.com
[ adapted to cpu_init_exception_handling(void) lacking FRED and LASS support ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 4c5e7f0fcd592801c9cc18f29f80fbee84eb8669 ]
On arm64 server, we found folio that get from migration entry isn't locked
in softleaf_to_folio(). This issue triggers when mTHP splitting and
zap_nonpresent_ptes() races, and the root cause is lack of memory barrier
in softleaf_to_folio(). The race is as follows:
CPU0 CPU1
deferred_split_scan() zap_nonpresent_ptes()
lock folio
split_folio()
unmap_folio()
change ptes to migration entries
__split_folio_to_order() softleaf_to_folio()
set flags(including PG_locked) for tail pages folio = pfn_folio(softleaf_to_pfn(entry))
smp_wmb() VM_WARN_ON_ONCE(!folio_test_locked(folio))
prep_compound_page() for tail pages
In __split_folio_to_order(), smp_wmb() guarantees page flags of tail pages
are visible before the tail page becomes non-compound. smp_wmb() should
be paired with smp_rmb() in softleaf_to_folio(), which is missed. As a
result, if zap_nonpresent_ptes() accesses migration entry that stores tail
pfn, softleaf_to_folio() may see the updated compound_head of tail page
before page->flags.
This issue will trigger VM_WARN_ON_ONCE() in pfn_swap_entry_folio()
because of the race between folio split and zap_nonpresent_ptes()
leading to a folio incorrectly undergoing modification without a folio
lock being held.
This is a BUG_ON() before commit 93976a20345b ("mm: eliminate further
swapops predicates"), which in merged in v6.19-rc1.
To fix it, add missing smp_rmb() if the softleaf entry is migration entry
in softleaf_to_folio() and softleaf_to_page().
[tujinjiang@huawei.com: update function name and comments]
Link: https://lkml.kernel.org/r/20260321075214.3305564-1-tujinjiang@huawei.com
Link: https://lkml.kernel.org/r/20260319012541.4158561-1-tujinjiang@huawei.com
Fixes: e9b61f19858a ("thp: reintroduce split_huge_page()")
Signed-off-by: Jinjiang Tu <tujinjiang@huawei.com>
Acked-by: David Hildenbrand (Arm) <david@kernel.org>
Reviewed-by: Lorenzo Stoakes (Oracle) <ljs@kernel.org>
Cc: Barry Song <baohua@kernel.org>
Cc: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Mike Rapoport <rppt@kernel.org>
Cc: Nanyong Sun <sunnanyong@huawei.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: Suren Baghdasaryan <surenb@google.com>
Cc: Vlastimil Babka <vbabka@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
[ adapted fix from leafops.h softleaf_to_page()/softleaf_to_folio() ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 1333eee56cdf3f0cf67c6ab4114c2c9e0a952026 ]
tcm_loop_target_reset() violates the SCSI EH contract: it returns SUCCESS
without draining any in-flight commands. The SCSI EH documentation
(scsi_eh.rst) requires that when a reset handler returns SUCCESS the driver
has made lower layers "forget about timed out scmds" and is ready for new
commands. Every other SCSI LLD (virtio_scsi, mpt3sas, ipr, scsi_debug,
mpi3mr) enforces this by draining or completing outstanding commands before
returning SUCCESS.
Because tcm_loop_target_reset() doesn't drain, the SCSI EH reuses in-flight
scsi_cmnd structures for recovery commands (e.g. TUR) while the target core
still has async completion work queued for the old se_cmd. The memset in
queuecommand zeroes se_lun and lun_ref_active, causing
transport_lun_remove_cmd() to skip its percpu_ref_put(). The leaked LUN
reference prevents transport_clear_lun_ref() from completing, hanging
configfs LUN unlink forever in D-state:
INFO: task rm:264 blocked for more than 122 seconds.
rm D 0 264 258 0x00004000
Call Trace:
__schedule+0x3d0/0x8e0
schedule+0x36/0xf0
transport_clear_lun_ref+0x78/0x90 [target_core_mod]
core_tpg_remove_lun+0x28/0xb0 [target_core_mod]
target_fabric_port_unlink+0x50/0x60 [target_core_mod]
configfs_unlink+0x156/0x1f0 [configfs]
vfs_unlink+0x109/0x290
do_unlinkat+0x1d5/0x2d0
Fix this by making tcm_loop_target_reset() actually drain commands:
1. Issue TMR_LUN_RESET via tcm_loop_issue_tmr() to drain all commands that
the target core knows about (those not yet CMD_T_COMPLETE).
2. Use blk_mq_tagset_busy_iter() to iterate all started requests and
flush_work() on each se_cmd — this drains any deferred completion work
for commands that already had CMD_T_COMPLETE set before the TMR (which
the TMR skips via __target_check_io_state()). This is the same pattern
used by mpi3mr, scsi_debug, and libsas to drain outstanding commands
during reset.
Fixes: e0eb5d38b732 ("scsi: target: tcm_loop: Use block cmd allocator for se_cmds")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-6
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Link: https://patch.msgid.link/27011aa34c8f6b1b94d2e3cf5655b6d037f53428.1773706803.git.josef@toxicpanda.com
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit c4ea7d8907cf72b259bf70bd8c2e791e1c4ff70f ]
If auxiliary_device_add() fails, add_adev() jumps to add_fail and calls
auxiliary_device_uninit(adev).
The auxiliary device has its release callback set to adev_release(),
which frees the containing struct mana_adev. Since adev is embedded in
struct mana_adev, the subsequent fall-through to init_fail and access
to adev->id may result in a use-after-free.
Fix this by saving the allocated auxiliary device id in a local
variable before calling auxiliary_device_add(), and use that saved id
in the cleanup path after auxiliary_device_uninit().
Fixes: a69839d4327d ("net: mana: Add support for auxiliary device")
Cc: stable@vger.kernel.org
Reviewed-by: Long Li <longli@microsoft.com>
Signed-off-by: Guangshuo Li <lgs201920130244@gmail.com>
Link: https://patch.msgid.link/20260323165730.945365-1-lgs201920130244@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit c4336a07eb6b2526dc2b62928b5104b41a7f81f5 ]
NETIF_F_IPV6_CSUM only advertises support for checksum offload of
packets without IPv6 extension headers. Packets with extension
headers must fall back onto software checksumming. Since TSO
depends on checksum offload, those must revert to GSO.
The below commit introduces that fallback. It always checks
network header length. For tunneled packets, the inner header length
must be checked instead. Extend the check accordingly.
A special case is tunneled packets without inner IP protocol. Such as
RFC 6951 SCTP in UDP. Those are not standard IPv6 followed by
transport header either, so also must revert to the software GSO path.
Cc: stable@vger.kernel.org
Fixes: 864e3396976e ("net: gso: Forbid IPv6 TSO with extensions on devices with only IPV6_CSUM")
Reported-by: Tangxin Xie <xietangxin@yeah.net>
Closes: https://lore.kernel.org/netdev/0414e7e2-9a1c-4d7c-a99d-b9039cf68f40@yeah.net/
Suggested-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Link: https://patch.msgid.link/20260320190148.2409107-1-willemdebruijn.kernel@gmail.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit 317e49358ebbf6390fa439ef3c142f9239dd25fb ]
The devm_free_irq() and devm_request_irq() functions should not be
executed in an atomic context.
During device suspend, all userspace processes and most kernel threads
are frozen. Additionally, we flush all tx/rx status, disable all macb
interrupts, and halt rx operations. Therefore, it is safe to split the
region protected by bp->lock into two independent sections, allowing
devm_free_irq() and devm_request_irq() to run in a non-atomic context.
This modification resolves the following lockdep warning:
BUG: sleeping function called from invalid context at kernel/locking/mutex.c:591
in_atomic(): 1, irqs_disabled(): 1, non_block: 0, pid: 501, name: rtcwake
preempt_count: 1, expected: 0
RCU nest depth: 1, expected: 0
7 locks held by rtcwake/501:
#0: ffff0008038c3408 (sb_writers#5){.+.+}-{0:0}, at: vfs_write+0xf8/0x368
#1: ffff0008049a5e88 (&of->mutex#2){+.+.}-{4:4}, at: kernfs_fop_write_iter+0xbc/0x1c8
#2: ffff00080098d588 (kn->active#70){.+.+}-{0:0}, at: kernfs_fop_write_iter+0xcc/0x1c8
#3: ffff800081c84888 (system_transition_mutex){+.+.}-{4:4}, at: pm_suspend+0x1ec/0x290
#4: ffff0008009ba0f8 (&dev->mutex){....}-{4:4}, at: device_suspend+0x118/0x4f0
#5: ffff800081d00458 (rcu_read_lock){....}-{1:3}, at: rcu_lock_acquire+0x4/0x48
#6: ffff0008031fb9e0 (&bp->lock){-.-.}-{3:3}, at: macb_suspend+0x144/0x558
irq event stamp: 8682
hardirqs last enabled at (8681): [<ffff8000813c7d7c>] _raw_spin_unlock_irqrestore+0x44/0x88
hardirqs last disabled at (8682): [<ffff8000813c7b58>] _raw_spin_lock_irqsave+0x38/0x98
softirqs last enabled at (7322): [<ffff8000800f1b4c>] handle_softirqs+0x52c/0x588
softirqs last disabled at (7317): [<ffff800080010310>] __do_softirq+0x20/0x2c
CPU: 1 UID: 0 PID: 501 Comm: rtcwake Not tainted 7.0.0-rc3-next-20260310-yocto-standard+ #125 PREEMPT
Hardware name: ZynqMP ZCU102 Rev1.1 (DT)
Call trace:
show_stack+0x24/0x38 (C)
__dump_stack+0x28/0x38
dump_stack_lvl+0x64/0x88
dump_stack+0x18/0x24
__might_resched+0x200/0x218
__might_sleep+0x38/0x98
__mutex_lock_common+0x7c/0x1378
mutex_lock_nested+0x38/0x50
free_irq+0x68/0x2b0
devm_irq_release+0x24/0x38
devres_release+0x40/0x80
devm_free_irq+0x48/0x88
macb_suspend+0x298/0x558
device_suspend+0x218/0x4f0
dpm_suspend+0x244/0x3a0
dpm_suspend_start+0x50/0x78
suspend_devices_and_enter+0xec/0x560
pm_suspend+0x194/0x290
state_store+0x110/0x158
kobj_attr_store+0x1c/0x30
sysfs_kf_write+0xa8/0xd0
kernfs_fop_write_iter+0x11c/0x1c8
vfs_write+0x248/0x368
ksys_write+0x7c/0xf8
__arm64_sys_write+0x28/0x40
invoke_syscall+0x4c/0xe8
el0_svc_common+0x98/0xf0
do_el0_svc+0x28/0x40
el0_svc+0x54/0x1e0
el0t_64_sync_handler+0x84/0x130
el0t_64_sync+0x198/0x1a0
Fixes: 558e35ccfe95 ("net: macb: WoL support for GEM type of Ethernet controller")
Cc: stable@vger.kernel.org
Reviewed-by: Théo Lebrun <theo.lebrun@bootlin.com>
Signed-off-by: Kevin Hao <haokexin@gmail.com>
Link: https://patch.msgid.link/20260318-macb-irq-v2-1-f1179768ab24@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ replaced `tmp` variable with direct `MACB_BIT(MAG)` ]
Signed-off-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
[ Upstream commit ffd32db8263d2d785a2c419486a450dc80693235 ]
ODR switch is done in 2 steps when FIFO is on : change the ODR register
value and acknowledge change when reading the FIFO ODR change flag.
When we are switching odr and turning buffer off just afterward, we are
losing the FIFO ODR change flag and ODR switch is blocked.
Fix the issue by force applying any waiting ODR change when turning
buffer off.
Fixes: ec74ae9fd37c ("iio: imu: inv_icm42600: add accurate timestamping")
Signed-off-by: Jean-Baptiste Maneyrol <jean-baptiste.maneyrol@tdk.com>
Cc: stable@vger.kernel.org
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 789b06f9f39cdc7e895bdab2c034e39c41c8f8d6 upstream.
Currently we execute `SET_NETDEV_DEV(dev, &priv->lowerdev->dev)` for
the virt_wifi net devices. However, unregistering a virt_wifi device in
netdev_run_todo() can happen together with the device referenced by
SET_NETDEV_DEV().
It can result in use-after-free during the ethtool operations performed
on a virt_wifi device that is currently being unregistered. Such a net
device can have the `dev.parent` field pointing to the freed memory,
but ethnl_ops_begin() calls `pm_runtime_get_sync(dev->dev.parent)`.
Let's remove SET_NETDEV_DEV for virt_wifi to avoid bugs like this:
==================================================================
BUG: KASAN: slab-use-after-free in __pm_runtime_resume+0xe2/0xf0
Read of size 2 at addr ffff88810cfc46f8 by task pm/606
Call Trace:
<TASK>
dump_stack_lvl+0x4d/0x70
print_report+0x170/0x4f3
? __pfx__raw_spin_lock_irqsave+0x10/0x10
kasan_report+0xda/0x110
? __pm_runtime_resume+0xe2/0xf0
? __pm_runtime_resume+0xe2/0xf0
__pm_runtime_resume+0xe2/0xf0
ethnl_ops_begin+0x49/0x270
ethnl_set_features+0x23c/0xab0
? __pfx_ethnl_set_features+0x10/0x10
? kvm_sched_clock_read+0x11/0x20
? local_clock_noinstr+0xf/0xf0
? local_clock+0x10/0x30
? kasan_save_track+0x25/0x60
? __kasan_kmalloc+0x7f/0x90
? genl_family_rcv_msg_attrs_parse.isra.0+0x150/0x2c0
genl_family_rcv_msg_doit+0x1e7/0x2c0
? __pfx_genl_family_rcv_msg_doit+0x10/0x10
? __pfx_cred_has_capability.isra.0+0x10/0x10
? stack_trace_save+0x8e/0xc0
genl_rcv_msg+0x411/0x660
? __pfx_genl_rcv_msg+0x10/0x10
? __pfx_ethnl_set_features+0x10/0x10
netlink_rcv_skb+0x121/0x380
? __pfx_genl_rcv_msg+0x10/0x10
? __pfx_netlink_rcv_skb+0x10/0x10
? __pfx_down_read+0x10/0x10
genl_rcv+0x23/0x30
netlink_unicast+0x60f/0x830
? __pfx_netlink_unicast+0x10/0x10
? __pfx___alloc_skb+0x10/0x10
netlink_sendmsg+0x6ea/0xbc0
? __pfx_netlink_sendmsg+0x10/0x10
? __futex_queue+0x10b/0x1f0
____sys_sendmsg+0x7a2/0x950
? copy_msghdr_from_user+0x26b/0x430
? __pfx_____sys_sendmsg+0x10/0x10
? __pfx_copy_msghdr_from_user+0x10/0x10
___sys_sendmsg+0xf8/0x180
? __pfx____sys_sendmsg+0x10/0x10
? __pfx_futex_wait+0x10/0x10
? fdget+0x2e4/0x4a0
__sys_sendmsg+0x11f/0x1c0
? __pfx___sys_sendmsg+0x10/0x10
do_syscall_64+0xe2/0x570
? exc_page_fault+0x66/0xb0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
</TASK>
This fix may be combined with another one in the ethtool subsystem:
https://lore.kernel.org/all/20260322075917.254874-1-alex.popov@linux.com/T/#u
Fixes: d43c65b05b848e0b ("ethtool: runtime-resume netdev parent in ethnl_ops_begin")
Cc: stable@vger.kernel.org
Signed-off-by: Alexander Popov <alex.popov@linux.com>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Reviewed-by: Breno Leitao <leitao@debian.org>
Link: https://patch.msgid.link/20260324224607.374327-1-alex.popov@linux.com
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 6e0e34d85cd46ceb37d16054e97a373a32770f6c upstream.
f_audio_complete() copies req->length bytes into a 4-byte stack
variable:
u32 data = 0;
memcpy(&data, req->buf, req->length);
req->length is derived from the host-controlled USB request path,
which can lead to a stack out-of-bounds write.
Validate req->actual against the expected payload size for the
supported control selectors and decode only the expected amount
of data.
This avoids copying a host-influenced length into a fixed-size
stack object.
Signed-off-by: Taegu Ha <hataegu0826@gmail.com>
Cc: stable <stable@kernel.org>
Link: https://patch.msgid.link/20260401191311.3604898-1-hataegu0826@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 8d8c68b1fc06ece60cf43e1306ff0f4ac121547e upstream.
The class/subclass/protocol options are suspectible to race conditions
as they can be accessed concurrently through configfs.
Use existing mutex to protect these options. This issue was identified
during code inspection.
Fixes: 73517cf49bd4 ("usb: gadget: add RNDIS configfs options for class/subclass/protocol")
Cc: stable@vger.kernel.org
Signed-off-by: Kuen-Han Tsai <khtsai@google.com>
Link: https://patch.msgid.link/20260320-usb-net-lifecycle-v1-2-4886b578161b@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit caa27923aacd8a5869207842f2ab1657c6c0c7bc upstream.
geth_alloc() increments the reference count, but geth_free() fails to
decrement it. This prevents the configuration of attributes via configfs
after unlinking the function.
Decrement the reference count in geth_free() to ensure proper cleanup.
Fixes: 02832e56f88a ("usb: gadget: f_subset: add configfs support")
Cc: stable@vger.kernel.org
Signed-off-by: Kuen-Han Tsai <khtsai@google.com>
Link: https://patch.msgid.link/20260320-usb-net-lifecycle-v1-1-4886b578161b@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit eba2936bbe6b752a31725a9eb5c674ecbf21ee7d upstream.
Commit b81ac4395bbe ("usb: gadget: uvc: allow for application to cleanly
shutdown") introduced two stages of synchronization waits totaling 1500ms
in uvc_function_unbind() to prevent several types of kernel panics.
However, this timing-based approach is insufficient during power
management (PM) transitions.
When the PM subsystem starts freezing user space processes, the
wait_event_interruptible_timeout() is aborted early, which allows the
unbind thread to proceed and nullify the gadget pointer
(cdev->gadget = NULL):
[ 814.123447][ T947] configfs-gadget.g1 gadget.0: uvc: uvc_function_unbind()
[ 814.178583][ T3173] PM: suspend entry (deep)
[ 814.192487][ T3173] Freezing user space processes
[ 814.197668][ T947] configfs-gadget.g1 gadget.0: uvc: uvc_function_unbind no clean disconnect, wait for release
When the PM subsystem resumes or aborts the suspend and tasks are
restarted, the V4L2 release path is executed and attempts to access the
already nullified gadget pointer, triggering a kernel panic:
[ 814.292597][ C0] PM: pm_system_irq_wakeup: 479 triggered dhdpcie_host_wake
[ 814.386727][ T3173] Restarting tasks ...
[ 814.403522][ T4558] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000030
[ 814.404021][ T4558] pc : usb_gadget_deactivate+0x14/0xf4
[ 814.404031][ T4558] lr : usb_function_deactivate+0x54/0x94
[ 814.404078][ T4558] Call trace:
[ 814.404080][ T4558] usb_gadget_deactivate+0x14/0xf4
[ 814.404083][ T4558] usb_function_deactivate+0x54/0x94
[ 814.404087][ T4558] uvc_function_disconnect+0x1c/0x5c
[ 814.404092][ T4558] uvc_v4l2_release+0x44/0xac
[ 814.404095][ T4558] v4l2_release+0xcc/0x130
Address the race condition and NULL pointer dereference by:
1. State Synchronization (flag + mutex)
Introduce a 'func_unbound' flag in struct uvc_device. This allows
uvc_function_disconnect() to safely skip accessing the nullified
cdev->gadget pointer. As suggested by Alan Stern, this flag is protected
by a new mutex (uvc->lock) to ensure proper memory ordering and prevent
instruction reordering or speculative loads. This mutex is also used to
protect 'func_connected' for consistent state management.
2. Explicit Synchronization (completion)
Use a completion to synchronize uvc_function_unbind() with the
uvc_vdev_release() callback. This prevents Use-After-Free (UAF) by
ensuring struct uvc_device is freed after all video device resources
are released.
Fixes: b81ac4395bbe ("usb: gadget: uvc: allow for application to cleanly shutdown")
Cc: stable <stable@kernel.org>
Suggested-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Jimmy Hu <hhhuuu@google.com>
Link: https://patch.msgid.link/20260320065427.1374555-1-hhhuuu@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit e1eabb072c75681f78312c484ccfffb7430f206e upstream.
A race condition between gether_disconnect() and eth_stop() leads to a
NULL pointer dereference. Specifically, if eth_stop() is triggered
concurrently while gether_disconnect() is tearing down the endpoints,
eth_stop() attempts to access the cleared endpoint descriptor, causing
the following NPE:
Unable to handle kernel NULL pointer dereference
Call trace:
__dwc3_gadget_ep_enable+0x60/0x788
dwc3_gadget_ep_enable+0x70/0xe4
usb_ep_enable+0x60/0x15c
eth_stop+0xb8/0x108
Because eth_stop() crashes while holding the dev->lock, the thread
running gether_disconnect() fails to acquire the same lock and spins
forever, resulting in a hardlockup:
Core - Debugging Information for Hardlockup core(7)
Call trace:
queued_spin_lock_slowpath+0x94/0x488
_raw_spin_lock+0x64/0x6c
gether_disconnect+0x19c/0x1e8
ncm_set_alt+0x68/0x1a0
composite_setup+0x6a0/0xc50
The root cause is that the clearing of dev->port_usb in
gether_disconnect() is delayed until the end of the function.
Move the clearing of dev->port_usb to the very beginning of
gether_disconnect() while holding dev->lock. This cuts off the link
immediately, ensuring eth_stop() will see dev->port_usb as NULL and
safely bail out.
Fixes: 2b3d942c4878 ("usb ethernet gadget: split out network core")
Cc: stable <stable@kernel.org>
Signed-off-by: Kuen-Han Tsai <khtsai@google.com>
Link: https://patch.msgid.link/20260311-gether-disconnect-npe-v1-1-454966adf7c7@google.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
With -fno-asynchronous-unwind-tables and --no-eh-frame-hdr (the default
of the linker), the GNU_EH_FRAME segment (specified by vdso.lds.S) is
empty. This is not valid, as the current DWARF specification mandates
the first byte of the EH frame to be the version number 1. It causes
some unwinders to complain, for example the ClickHouse query profiler
spams the log with messages:
clickhouse-server[365854]: libunwind: unsupported .eh_frame_hdr
version: 127 at 7ffffffb0000
Here "127" is just the byte located at the p_vaddr (0, i.e. the
beginning of the vDSO) of the empty GNU_EH_FRAME segment. Cross-
checking with /proc/365854/maps has also proven 7ffffffb0000 is the
start of vDSO in the process VM image.
In LoongArch the -fno-asynchronous-unwind-tables option seems just a
MIPS legacy, and MIPS only uses this option to satisfy the MIPS-specific
"genvdso" program, per the commit cfd75c2db17e ("MIPS: VDSO: Explicitly
use -fno-asynchronous-unwind-tables"). IIRC it indicates some inherent
limitation of the MIPS ELF ABI and has nothing to do with LoongArch. So
we can simply flip it over to -fasynchronous-unwind-tables and pass
--eh-frame-hdr for linking the vDSO, allowing the profilers to unwind the
stack for statistics even if the sample point is taken when the PC is in
the vDSO.
However simply adjusting the options above would exploit an issue: when
the libgcc unwinder saw the invalid GNU_EH_FRAME segment, it silently
falled back to a machine-specific routine to match the code pattern of
rt_sigreturn() and extract the registers saved in the sigframe if the
code pattern is matched. As unwinding from signal handlers is vital for
libgcc to support pthread cancellation etc., the fall-back routine had
been silently keeping the LoongArch Linux systems functioning since
Linux 5.19. But when we start to emit GNU_EH_FRAME with the correct
format, fall-back routine will no longer be used and libgcc will fail
to unwind the sigframe, and unwinding from signal handlers will no
longer work, causing dozens of glibc test failures. To make it possible
to unwind from signal handlers again, it's necessary to code the unwind
info in __vdso_rt_sigreturn via .cfi_* directives.
The offsets in the .cfi_* directives depend on the layout of struct
sigframe, notably the offset of sigcontext in the sigframe. To use the
offset in the assembly file, factor out struct sigframe into a header to
allow asm-offsets.c to output the offset for assembly.
To work around a long-term issue in the libgcc unwinder (the pc is
unconditionally substracted by 1: doing so is technically incorrect for
a signal frame), a nop instruction is included with the two real
instructions in __vdso_rt_sigreturn in the same FDE PC range. The same
hack has been used on x86 for a long time.
(cherry picked from commit e4878c37f6679fdea91b27a0f4e60a871f0b7bad)
Cc: stable@vger.kernel.org
Fixes: c6b99bed6b8f ("LoongArch: Add VDSO and VSYSCALL support")
Signed-off-by: Xi Ruoyao <xry111@xry111.site>
Signed-off-by: Huacai Chen <chenhuacai@loongson.cn>
Signed-off-by: Xi Ruoyao <xry111@xry111.site>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit 557c024ca7250bb65ae60f16c02074106c2f197b ]
A fuzzer test introduced corruption that ends up with a depth of 0 in
dir_e_read(), causing an undefined shift by 32 at:
index = hash >> (32 - dip->i_depth);
As calculated in an open-coded way in dir_make_exhash(), the minimum
depth for an exhash directory is ilog2(sdp->sd_hash_ptrs) and 0 is
invalid as sdp->sd_hash_ptrs is fixed as sdp->bsize / 16 at mount time.
So we can avoid the undefined behaviour by checking for depth values
lower than the minimum in gfs2_dinode_in(). Values greater than the
maximum are already being checked for there.
Also switch the calculation in dir_make_exhash() to use ilog2() to
clarify how the depth is calculated.
Tested with the syzkaller repro.c and xfstests '-g quick'.
Reported-by: syzbot+4708579bb230a0582a57@syzkaller.appspotmail.com
Signed-off-by: Andrew Price <anprice@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit 10398ef57aa189153406c110f5957145030f08fe ]
gfs2_consist_inode() logs an error message with the source file and line
number. When we jump before calling it, the line number becomes less
useful as it no longer relates to the source of the error. To aid
troubleshooting, replace the gotos with the gfs2_consist_inode() calls
so that the error messages are more informative.
Signed-off-by: Andrew Price <anprice@redhat.com>
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Ruohan Lan <ruohanlan@aliyun.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit f8da41de0bff9eb1d774a7253da0c9f637c4470a ]
If we fail to create an inline extent due to -ENOSPC, we will attempt to
go through the normal COW path, reserve an extent, create an ordered
extent, etc. However we were always freeing the reserved qgroup data,
which is wrong since we will use data. Fix this by freeing the reserved
qgroup data in __cow_file_range_inline() only if we are not doing the
fallback (ret is <= 0).
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
[ Upstream commit 0bb067ca64e35536f1f5d9ef6aaafc40f4833623 ]
Inside function __cow_file_range_inline() since the inlined data no
longer take any data space, we need to free up the reserved space.
However the code is still using the old page size == sector size
assumption, and will not handle subpage case well.
Thankfully it is not going to cause any problems because we have two extra
safe nets:
- Inline data extents creation is disabled for sector size < page size
cases for now
But it won't stay that for long.
- btrfs_qgroup_free_data() will only clear ranges which have been already
reserved
So even if we pass a range larger than what we need, it should still
be fine, especially there is only reserved space for a single block at
file offset 0 of an inline data extent.
But just for the sake of consistency, fix the call site to use
sectorsize instead of page size.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Stable-dep-of: f8da41de0bff ("btrfs: do not free data reservation in fallback from inline due to -ENOSPC")
Signed-off-by: Sasha Levin <sashal@kernel.org>
|
|
partial transfer
commit f50200dd44125e445a6164e88c217472fa79cdbc upstream.
When a gadget request is only partially transferred in transfer()
because the per-frame bandwidth budget is exhausted, the loop advances
to the next queued request. If that next request is a zero-length
packet (ZLP), len evaluates to zero and the code takes the
unlikely(len == 0) path, which sets is_short = 1. This bypasses the
bandwidth guard ("limit < ep->ep.maxpacket && limit < len") that
lives in the else branch and would otherwise break out of the loop for
non-zero requests. The is_short path then completes the URB before all
data from the first request has been transferred.
Reproducer (bulk IN, high speed):
Device side (FunctionFS with Linux AIO):
1. Queue a 65024-byte write via io_submit (127 * 512, i.e. a
multiple of the HS bulk max packet size).
2. Immediately queue a zero-length write (ZLP) via io_submit.
Host side:
3. Submit a 65536-byte bulk IN URB.
Expected: URB completes with actual_length = 65024.
Actual: URB completes with actual_length = 53248, losing 11776
bytes that leak into subsequent URBs.
At high speed the per-frame budget is 53248 bytes (512 * 13 * 8).
The 65024-byte request exhausts this budget after 53248 bytes, leaving
the request incomplete (req->req.actual < req->req.length). Neither
the request nor the URB is finished, and rescan is 0, so the loop
advances to the ZLP. For the ZLP, dev_len = 0, so len = min(12288, 0)
= 0, taking the unlikely(len == 0) path and setting is_short = 1.
The is_short handler then sets *status = 0, completing the URB with
only 53248 of the expected 65024 bytes.
Fix this by breaking out of the loop when the current request has
remaining data (req->req.actual < req->req.length). The request
resumes on the next timer tick, preserving correct data ordering.
Signed-off-by: Sebastian Urban <surban@surban.net>
Cc: stable <stable@kernel.org>
Reviewed-by: Alan Stern <stern@rowland.harvard.edu>
Link: https://patch.msgid.link/20260315151045.1155850-1-surban@surban.net
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 2ca9e46f8f1f5a297eb0ac83f79d35d5b3a02541 upstream.
This fixes an error in synchronization in the dummy-hcd driver. The
error has a somewhat involved history. The synchronization mechanism
was introduced by commit 7dbd8f4cabd9 ("USB: dummy-hcd: Fix erroneous
synchronization change"), which added an emulated "interrupts enabled"
flag together with code emulating synchronize_irq() (it waits until
all current handler callbacks have returned).
But the emulated interrupt-disable occurred too late, after the driver
containing the handler callback routines had been told that it was
unbound and no more callbacks would occur. Commit 4a5d797a9f9c ("usb:
gadget: dummy_hcd: fix gpf in gadget_setup") tried to fix this by
moving the synchronize_irq() emulation code from dummy_stop() to
dummy_pullup(), which runs before the unbind callback.
There still were races, though, because the emulated interrupt-disable
still occurred too late. It couldn't be moved to dummy_pullup(),
because that routine can be called for reasons other than an impending
unbind. Therefore commits 7dc0c55e9f30 ("USB: UDC core: Add
udc_async_callbacks gadget op") and 04145a03db9d ("USB: UDC: Implement
udc_async_callbacks in dummy-hcd") added an API allowing the UDC core
to tell dummy-hcd exactly when emulated interrupts and their callbacks
should be disabled.
That brings us to the current state of things, which is still wrong
because the emulated synchronize_irq() occurs before the emulated
interrupt-disable! That's no good, beause it means that more emulated
interrupts can occur after the synchronize_irq() emulation has run,
leading to the possibility that a callback handler may be running when
the gadget driver is unbound.
To fix this, we have to move the synchronize_irq() emulation code yet
again, to the dummy_udc_async_callbacks() routine, which takes care of
enabling and disabling emulated interrupt requests. The
synchronization will now run immediately after emulated interrupts are
disabled, which is where it belongs.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Fixes: 04145a03db9d ("USB: UDC: Implement udc_async_callbacks in dummy-hcd")
Cc: stable <stable@kernel.org>
Link: https://patch.msgid.link/c7bc93fe-4241-4d04-bd56-27c12ba35c97@rowland.harvard.edu
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
commit 616a63ff495df12863692ab3f9f7b84e3fa7a66d upstream.
Syzbot testing was able to provoke an addressing exception and crash
in the usb_gadget_udc_reset() routine in
drivers/usb/gadgets/udc/core.c, resulting from the fact that the
routine was called with a second ("driver") argument of NULL. The bad
caller was set_link_state() in dummy_hcd.c, and the problem arose
because of a race between a USB reset and driver unbind.
These sorts of races were not supposed to be possible; commit
7dbd8f4cabd9 ("USB: dummy-hcd: Fix erroneous synchronization change"),
along with a few followup commits, was written specifically to prevent
them. As it turns out, there are (at least) two errors remaining in
the code. Another patch will address the second error; this one is
concerned with the first.
The error responsible for the syzbot crash occurred because the
stop_activity() routine will sometimes drop and then re-acquire the
dum->lock spinlock. A call to stop_activity() occurs in
set_link_state() when handling an emulated USB reset, after the test
of dum->ints_enabled and before the increment of dum->callback_usage.
This allowed another thread (doing a driver unbind) to sneak in and
grab the spinlock, and then clear dum->ints_enabled and dum->driver.
Normally this other thread would have to wait for dum->callback_usage
to go down to 0 before it would clear dum->driver, but in this case it
didn't have to wait since dum->callback_usage had not yet been
incremented.
The fix is to increment dum->callback_usage _before_ calling
stop_activity() instead of after. Then the thread doi |
