Exercise various mmap(), munmap() and mremap() invocations, which might
cause a perf buffer mapping to be split or truncated.
To avoid hard coding the perf event and having dependencies on
architectures and configuration options, scan through event types in sysfs
and try to open them. On success, try to mmap() and if that succeeds try to
mmap() the AUX buffer.
In case that no AUX buffer supporting event is found, only test the base
buffer mapping. If no mappable event is found or permissions are not
sufficient, skip the tests.
Reserve a PROT_NONE region for both rb and aux tests to allow testing the
case where mremap unmaps beyond the end of a mapped VMA to prevent it from
unmapping unrelated mappings.
Signed-off-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com> Co-developed-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The perf mmap code is careful about mmap()'ing the user page with the
ringbuffer and additionally the auxiliary buffer, when the event supports
it. Once the first mapping is established, subsequent mapping have to use
the same offset and the same size in both cases. The reference counting for
the ringbuffer and the auxiliary buffer depends on this being correct.
Though perf does not prevent that a related mapping is split via mmap(2),
munmap(2) or mremap(2). A split of a VMA results in perf_mmap_open() calls,
which take reference counts, but then the subsequent perf_mmap_close()
calls are not longer fulfilling the offset and size checks. This leads to
reference count leaks.
As perf already has the requirement for subsequent mappings to match the
initial mapping, the obvious consequence is that VMA splits, caused by
resizing of a mapping or partial unmapping, have to be prevented.
Implement the vm_operations_struct::may_split() callback and return
unconditionally -EINVAL.
That ensures that the mapping offsets and sizes cannot be changed after the
fact. Remapping to a different fixed address with the same size is still
possible as it takes the references for the new mapping and drops those of
the old mapping.
Fixes: 45bfb2e50471 ("perf/core: Add AUX area to ring buffer for raw data streams") Reported-by: zdi-disclosures@trendmicro.com # ZDI-CAN-27504 Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com> Acked-by: Arnaldo Carvalho de Melo <acme@redhat.com> Acked-by: Vlastimil Babka <vbabka@suse.cz> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
After successful allocation of a buffer or a successful attachment to an
existing buffer perf_mmap() tries to map the buffer read only into the page
table. If that fails, the already set up page table entries are zapped, but
the other perf specific side effects of that failure are not handled. The
calling code just cleans up the VMA and does not invoke perf_mmap_close().
This leaks reference counts, corrupts user->vm accounting and also results
in an unbalanced invocation of event::event_mapped().
Cure this by moving the event::event_mapped() invocation before the
map_range() call so that on map_range() failure perf_mmap_close() can be
invoked without causing an unbalanced event::event_unmapped() call.
perf_mmap_close() undoes the reference counts and eventually frees buffers.
Fixes: b709eb872e19 ("perf/core: map pages in advance") Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When perf_mmap() fails to allocate a buffer, it still invokes the
event_mapped() callback of the related event. On X86 this might increase
the perf_rdpmc_allowed reference counter. But nothing undoes this as
perf_mmap_close() is never called in this case, which causes another
reference count leak.
Return early on failure to prevent that.
Fixes: 1e0fb9ec679c ("perf/core: Add pmu callbacks to track event mapping and unmapping") Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Failure of the AUX buffer allocation leaks the reference count.
Set the reference count to 1 only when the allocation succeeds.
Fixes: 45bfb2e50471 ("perf/core: Add AUX area to ring buffer for raw data streams") Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
A recent overhaul sets the return value to 0 unconditionally after the
allocations, which causes reference count leaks and corrupts the user->vm
accounting.
Preserve the AUX buffer allocation failure return value, so that the
subsequent code works correctly.
Fixes: 0983593f32c4 ("perf/core: Lift event->mmap_mutex in perf_mmap()") Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Reviewed-by: Lorenzo Stoakes <lorenzo.stoakes@oracle.com> Cc: stable@vger.kernel.org Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Scott Mayhew discovered a security exploit in NFS over TLS in
tls_alert_recv() due to its assumption it can read data from
the msg iterator's kvec..
kTLS implementation splits TLS non-data record payload between
the control message buffer (which includes the type such as TLS
aler or TLS cipher change) and the rest of the payload (say TLS
alert's level/description) which goes into the msg payload buffer.
This patch proposes to rework how control messages are setup and
used by sock_recvmsg().
If no control message structure is setup, kTLS layer will read and
process TLS data record types. As soon as it encounters a TLS control
message, it would return an error. At that point, NFS can setup a
kvec backed msg buffer and read in the control message such as a
TLS alert. Msg iterator can advance the kvec pointer as a part of
the copy process thus we need to revert the iterator before calling
into the tls_alert_recv.
Reported-by: Scott Mayhew <smayhew@redhat.com> Fixes: 5e052dda121e ("SUNRPC: Recognize control messages in server-side TCP socket code") Suggested-by: Trond Myklebust <trondmy@hammerspace.com> Cc: stable@vger.kernel.org Signed-off-by: Olga Kornievskaia <okorniev@redhat.com> Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If two calls to nfsd_open_local_fh() race and both successfully call
nfsd_file_acquire_local(), they will both get an extra reference to the
net to accompany the file reference stored in *pnf.
One of them will fail to store (using xchg()) the file reference in
*pnf and will drop that reference but WON'T drop the accompanying
reference to the net. This leak means that when the nfs server is shut
down it will hang in nfsd_shutdown_net() waiting for
&nn->nfsd_net_free_done.
This patch adds the missing nfsd_net_put().
Reported-by: Mike Snitzer <snitzer@kernel.org> Fixes: e6f7e1487ab5 ("nfs_localio: simplify interface to nfsd for getting nfsd_file") Cc: stable@vger.kernel.org Signed-off-by: NeilBrown <neil@brown.name> Tested-by: Mike Snitzer <snitzer@kernel.org> Reviewed-by: Mike Snitzer <snitzer@kernel.org> Reviewed-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Clients will typically precede a DELEGRETURN for a delegation with
delegated timestamp with a SETATTR to set the timestamps on the server
to match what the client has.
knfsd implements this by using the nfsd_setattr() infrastructure, which
will set ATTR_CTIME on any update that goes to notify_change(). This is
problematic as it means that the client will get a spurious ctime
update when updating the atime.
POSIX unfortunately doesn't phrase it succinctly, but updating the atime
due to reads should not update the ctime. In this case, the client is
sending a SETATTR to update the atime on the server to match its latest
value. The ctime should not be advanced in this case as that would
incorrectly indicate a change to the inode.
Fix this by not implicitly setting ATTR_CTIME when ATTR_DELEG is set in
__nfsd_setattr(). The decoder for FATTR4_WORD2_TIME_DELEG_MODIFY already
sets ATTR_CTIME, so this is sufficient to make it skip setting the ctime
on atime-only updates.
Fixes: 7e13f4f8d27d ("nfsd: handle delegated timestamps in SETATTR") Cc: stable@vger.kernel.org Signed-off-by: Jeff Layton <jlayton@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Together with the RAPL MSRs, there are more MSRs gone on DMR, including
PLR (Perf Limit Reasons), and IRTL (Package cstate Interrupt Response
Time Limit) MSRs. The configurable TDP info should also be retrieved
from TPMI based Intel Speed Select Technology feature.
Remove the access of these MSRs for DMR. Improve the DMR platform
feature table to make it more readable at the same time.
Fixes: 83075bd59de2 ("tools/power turbostat: Add initial support for DMR") Signed-off-by: Zhang Rui <rui.zhang@intel.com> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Cc: Steve French <smfrench@gmail.com> Cc: Tom Talpey <tom@talpey.com> Cc: Long Li <longli@microsoft.com> Cc: linux-cifs@vger.kernel.org Cc: samba-technical@lists.samba.org Fixes: f198186aa9bb ("CIFS: SMBD: Establish SMB Direct connection") Signed-off-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit 528589947c180 ("nvmet: initialize discovery subsys after debugfs
is initialized") changed nvmet_init() to initialize nvme discovery after
"nvmet" debugfs directory is initialized. The change broke nvmet_exit()
because discovery subsystem now depends on debugfs. Debugfs should be
destroyed after discovery subsystem. Fix nvmet_exit() to do that.
Reported-by: Yi Zhang <yi.zhang@redhat.com> Closes: https://lore.kernel.org/all/CAHj4cs96AfFQpyDKF_MdfJsnOEo=2V7dQgqjFv+k3t7H-=yGhA@mail.gmail.com/ Fixes: 528589947c180 ("nvmet: initialize discovery subsys after debugfs is initialized") Signed-off-by: Mohamed Khalfella <mkhalfella@purestorage.com> Reviewed-by: Hannes Reinecke <hare@suse.de> Reviewed-by: Daniel Wagner <dwagner@suse.de> Link: https://lore.kernel.org/r/20250807053507.2794335-1-mkhalfella@purestorage.com Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Sasha Levin <sashal@kernel.org>
Calling enqueue_reassembly() and wake_up_interruptible(&info->wait_reassembly_queue)
or put_receive_buffer() means the response/data_transfer pointer might
get re-used by another thread, which means these should be
the last operations before calling return.
Cc: Steve French <smfrench@gmail.com> Cc: Tom Talpey <tom@talpey.com> Cc: Long Li <longli@microsoft.com> Cc: linux-cifs@vger.kernel.org Cc: samba-technical@lists.samba.org Fixes: f198186aa9bb ("CIFS: SMBD: Establish SMB Direct connection") Signed-off-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
We should call put_receive_buffer() before waking up the callers.
For the internal error case of response->type being unexpected,
we now also call smbd_disconnect_rdma_connection() instead
of not waking up the callers at all.
Note that the SMBD_TRANSFER_DATA case still has problems,
which will be addressed in the next commit in order to make
it easier to review this one.
Cc: Steve French <smfrench@gmail.com> Cc: Tom Talpey <tom@talpey.com> Cc: Long Li <longli@microsoft.com> Cc: linux-cifs@vger.kernel.org Cc: samba-technical@lists.samba.org Fixes: f198186aa9bb ("CIFS: SMBD: Establish SMB Direct connection") Signed-off-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
In case of failures either ib_dma_map_single() might not be called yet
or ib_dma_unmap_single() was already called.
We should make sure put_receive_buffer() only calls
ib_dma_unmap_single() if needed.
Cc: Steve French <smfrench@gmail.com> Cc: Tom Talpey <tom@talpey.com> Cc: Long Li <longli@microsoft.com> Cc: linux-cifs@vger.kernel.org Cc: samba-technical@lists.samba.org Fixes: f198186aa9bb ("CIFS: SMBD: Establish SMB Direct connection") Signed-off-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
There's no need to maintain two lists, we can just
have a single list of receive buffers, which are free to use.
It just added unneeded complexity and resulted in
ib_dma_unmap_single() not being called from recv_done()
for empty keepalive packets.
Cc: Steve French <smfrench@gmail.com> Cc: Tom Talpey <tom@talpey.com> Cc: Long Li <longli@microsoft.com> Cc: linux-cifs@vger.kernel.org Cc: samba-technical@lists.samba.org Fixes: f198186aa9bb ("CIFS: SMBD: Establish SMB Direct connection") Signed-off-by: Stefan Metzmacher <metze@samba.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Calling enqueue_reassembly() and wake_up_interruptible(&t->wait_reassembly_queue)
or put_receive_buffer() means the recvmsg/data_transfer pointer might
get re-used by another thread, which means these should be
the last operations before calling return.
Cc: Namjae Jeon <linkinjeon@kernel.org> Cc: Steve French <smfrench@gmail.com> Cc: Tom Talpey <tom@talpey.com> Cc: linux-cifs@vger.kernel.org Cc: samba-technical@lists.samba.org Fixes: 0626e6641f6b ("cifsd: add server handler for central processing and tranport layers") Signed-off-by: Stefan Metzmacher <metze@samba.org> Acked-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
We should call put_recvmsg() before smb_direct_disconnect_rdma_connection()
in order to call it before waking up the callers.
In all error cases we should call smb_direct_disconnect_rdma_connection()
in order to avoid stale connections.
Cc: Namjae Jeon <linkinjeon@kernel.org> Cc: Steve French <smfrench@gmail.com> Cc: Tom Talpey <tom@talpey.com> Cc: linux-cifs@vger.kernel.org Cc: samba-technical@lists.samba.org Fixes: 0626e6641f6b ("cifsd: add server handler for central processing and tranport layers") Signed-off-by: Stefan Metzmacher <metze@samba.org> Acked-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
In case of failures either ib_dma_map_single() might not be called yet
or ib_dma_unmap_single() was already called.
We should make sure put_recvmsg() only calls ib_dma_unmap_single() if needed.
Cc: Namjae Jeon <linkinjeon@kernel.org> Cc: Steve French <smfrench@gmail.com> Cc: Tom Talpey <tom@talpey.com> Cc: linux-cifs@vger.kernel.org Cc: samba-technical@lists.samba.org Fixes: 0626e6641f6b ("cifsd: add server handler for central processing and tranport layers") Signed-off-by: Stefan Metzmacher <metze@samba.org> Acked-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
There's no need to maintain two lists, we can just
have a single list of receive buffers, which are free to use.
Cc: Steve French <smfrench@gmail.com> Cc: Tom Talpey <tom@talpey.com> Cc: linux-cifs@vger.kernel.org Cc: samba-technical@lists.samba.org Fixes: 0626e6641f6b ("cifsd: add server handler for central processing and tranport layers") Signed-off-by: Stefan Metzmacher <metze@samba.org> Acked-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Steve French <stfrench@microsoft.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The hda-sdw-bpt code links against the soundwire driver, but that fails when
trying to link from built-in code into loadable module:
x86_64-linux-ld: vmlinux.o: in function `intel_ace2x_bpt_close_stream.isra.0':
intel_ace2x.c:(.text+0x137a531): undefined reference to `hda_sdw_bpt_close'
x86_64-linux-ld: vmlinux.o: in function `intel_ace2x_bpt_send_async':
intel_ace2x.c:(.text+0x137aa45): undefined reference to `hda_sdw_bpt_open'
x86_64-linux-ld: intel_ace2x.c:(.text+0x137ab67): undefined reference to `hda_sdw_bpt_close'
x86_64-linux-ld: intel_ace2x.c:(.text+0x137ac30): undefined reference to `hda_sdw_bpt_send_async'
x86_64-linux-ld: vmlinux.o: in function `intel_ace2x_bpt_wait':
intel_ace2x.c:(.text+0x137aced): undefined reference to `hda_sdw_bpt_wait'
Ensure that both SOUNDWIRE_INTEL and SND_SOF_SOF_HDA_SDW_BPT are selected
at the same time by SND_SOC_SOF_INTEL_LNL, and that this happens even if
SND_SOC_SOF_INTEL_SOUNDWIRE is a loadable module but SND_SOC_SOF_INTEL_LNL
is built-in.
This follows the same logic as commit c5a61db9bf89 ("ASoC: SOF: fix
intel-soundwire link failure").
Fixes: 5d5cb86fb46e ("ASoC: SOF: Intel: hda-sdw-bpt: add helpers for SoundWire BPT DMA") Signed-off-by: Arnd Bergmann <arnd@arndb.de> Reviewed-by: Bard Liao <yung-chuan.liao@linux.intel.com> Link: https://patch.msgid.link/20250805160451.4004602-1-arnd@kernel.org Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Compile-testing IMX_MU_MSI on x86 without PCI_MSI support results in a
build failure:
drivers/gpio/gpio-sprd.c:8:
include/linux/gpio/driver.h:41:33: error: field 'msiinfo' has incomplete type
drivers/iommu/iommufd/viommu.c:4:
include/linux/msi.h:528:33: error: field 'alloc_info' has incomplete type
Tighten the dependency further to only allow compile testing on Arm.
This could be refined further to allow certain x86 configs.
This was submitted before to address a different build failure, which was
fixed differently, but the problem has now returned in a different form.
emac_rx_packet() is a common function for handling traffic
for both xdp and non-xdp use cases. Use common logic for
handling skb with or without xdp to prevent any incorrect
packet processing. This patch fixes ping working with
XDP_PASS for icssg driver.
In order for the wait in nfs_uuid_put() to be safe, it is necessary to
ensure that nfs_uuid_add_file() doesn't add a new entry once the
nfs_uuid->net has been NULLed out.
Also fix up the wake_up_var_locked() / wait_var_event_spinlock() to both
use the nfs_uuid address, since nfl, and &nfl->uuid could be used elsewhere.
Acked-by: Mike Snitzer <snitzer@kernel.org> Tested-by: Mike Snitzer <snitzer@kernel.org> Link: https://lore.kernel.org/all/175262893035.2234665.1735173020338594784@noble.neil.brown.name/ Fixes: 21fb44034695 ("nfs_localio: protect race between nfs_uuid_put() and nfs_close_local_fh()") Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Wrap copying of drop stats on TX path from fbd->hw_stats by the
hw_stats_lock. Currently, it is being performed outside the lock and
another thread accessing fbd->hw_stats can lead to inconsistencies.
Fixes: 5f8bd2ce8269 ("eth: fbnic: add support for TMI stats") Signed-off-by: Mohsin Bashir <mohsin.bashr@gmail.com> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://patch.msgid.link/20250802024636.679317-3-mohsin.bashr@gmail.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Alex added page bias of LONG_MAX, which is admittedly quite
a clever way of catching overflows of the pp ref count.
The page pool code was "optimized" to leave the ref at 1
for freed pages so it can't catch basic bugs by itself any more.
(Something we should probably address under DEBUG_NET...)
Unfortunately for fbnic since commit f7dc3248dcfb ("skbuff: Optimization
of SKB coalescing for page pool") core _may_ actually take two extra
pp refcounts, if one of them is returned before driver gives up the bias
the ret < 0 check in page_pool_unref_netmem() will trigger.
While at it add a FBNIC_ to the name of the driver constant.
Make vmem_pte_alloc() consistent by always allocating page table of
PAGE_SIZE granularity, regardless of whether page_table_alloc() (with
slab) or memblock_alloc() is used. This ensures page table can be fully
freed when the corresponding page table entries are removed.
Fixes: d08d4e7cd6bf ("s390/mm: use full 4KB page for 2KB PTE") Reviewed-by: Heiko Carstens <hca@linux.ibm.com> Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com> Signed-off-by: Sumanth Korikkar <sumanthk@linux.ibm.com> Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
TCA_MQPRIO_TC_ENTRY_INDEX is validated using
NLA_POLICY_MAX(NLA_U32, TC_QOPT_MAX_QUEUE), which allows the value
TC_QOPT_MAX_QUEUE (16). This leads to a 4-byte out-of-bounds stack
write in the fp[] array, which only has room for 16 elements (0–15).
Fix this by changing the policy to allow only up to TC_QOPT_MAX_QUEUE - 1.
Fixes: f62af20bed2d ("net/sched: mqprio: allow per-TC user input of FP adminStatus") Reviewed-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Maher Azzouzi <maherazz04@gmail.com> Reviewed-by: Vladimir Oltean <vladimir.oltean@nxp.com> Link: https://patch.msgid.link/20250802001857.2702497-1-kuba@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Using devm_*() [here] is completely wrong, because this is called
from mdiobus_register_device(). This is not the probe function
for the device, and thus there is no code to trigger the release of
the resource on unregistration.
Moreover, when the mdiodev is eventually probed, if the driver fails
or the driver is unbound, the GPIO will be released, but a reference
will be left behind.
Using devm* with a struct device that is *not* currently being probed
is fundamentally wrong - an abuse of devm.
There are multiple drivers that use the qualcomm mdt loader, but they
have conflicting ideas of how to deal with that dependency when compile-testing
for non-qualcomm targets:
IPA only enables the MDT loader when the kernel config includes ARCH_QCOM,
but the newly added ath12k support always enables it, which leads to a
link failure with the combination of IPA=y and ATH12K=m:
aarch64-linux-ld: drivers/net/ipa/ipa_main.o: in function `ipa_firmware_load':
ipa_main.c:(.text.unlikely+0x134): undefined reference to `qcom_mdt_load
The ATH12K method seems more reliable here, so change IPA over to do the same
thing.
CI hit a UaF in fbnic in the AF_XDP portion of the queues.py test.
The UaF is in the __sk_mark_napi_id_once() call in xsk_bind(),
NAPI has been freed. Looks like the device failed to open earlier,
and we lack clearing the NAPI pointer from the queue.
Fixes: 557d02238e05 ("eth: fbnic: centralize the queue count and NAPI<>queue setting") Reviewed-by: Alexander Duyck <alexanderduyck@fb.com> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://patch.msgid.link/20250728163129.117360-1-kuba@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Hogan reported a vector setup race, which overwrites the interrupt
descriptor in the per CPU vector array resulting in a disfunctional device.
CPU0 CPU1
interrupt is raised in APIC IRR
but not handled
free_irq()
per_cpu(vector_irq, CPU1)[vector] = VECTOR_SHUTDOWN;
request_irq() common_interrupt()
d = this_cpu_read(vector_irq[vector]);
per_cpu(vector_irq, CPU1)[vector] = desc;
if (d == VECTOR_SHUTDOWN)
this_cpu_write(vector_irq[vector], VECTOR_UNUSED);
free_irq() cannot observe the pending vector in the CPU1 APIC as there is
no way to query the remote CPUs APIC IRR.
This requires that request_irq() uses the same vector/CPU as the one which
was freed, but this also can be triggered by a spurious interrupt.
Interestingly enough this problem managed to be hidden for more than a
decade.
Prevent this by reevaluating vector_irq under the vector lock, which is
held by the interrupt activation code when vector_irq is updated.
To avoid ifdeffery or IS_ENABLED() nonsense, move the
[un]lock_vector_lock() declarations out under the
CONFIG_IRQ_DOMAIN_HIERARCHY guard as it's only provided when
CONFIG_X86_LOCAL_APIC=y.
The current CONFIG_IRQ_DOMAIN_HIERARCHY guard is selected by
CONFIG_X86_LOCAL_APIC, but can also be selected by other parts of the
Kconfig system, which makes 32-bit UP builds with CONFIG_X86_LOCAL_APIC=n
fail.
Can we just get rid of this !APIC nonsense once and forever?
Fixes: 9345005f4eed ("x86/irq: Fix do_IRQ() interrupt warning for cpu hotplug retriggered irqs") Reported-by: Hogan Wang <hogan.wang@huawei.com> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Tested-by: Hogan Wang <hogan.wang@huawei.com> Link: https://lore.kernel.org/all/draft-87ikjhrhhh.ffs@tglx Signed-off-by: Sasha Levin <sashal@kernel.org>
We can't let restart worker run once device is removed, since other
data that it might want to access could be already released.
Explicitly disable worker as part of device cleanup action.
Fixes: a4d1c5d0b99b ("drm/xe/pf: Move VFs reprovisioning to worker") Signed-off-by: Michal Wajdeczko <michal.wajdeczko@intel.com> Reviewed-by: Piotr Piórkowski <piotr.piorkowski@intel.com> Cc: Jonathan Cavitt <jonathan.cavitt@intel.com> Link: https://lore.kernel.org/r/20250801142822.180530-2-michal.wajdeczko@intel.com
(cherry picked from commit a424353937c24554bb242a6582ed8f018b4a411c) Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
A security exploit was discovered in NFS over TLS in tls_alert_recv
due to its assumption that there is valid data in the msghdr's
iterator's kvec.
Instead, this patch proposes the rework how control messages are
setup and used by sock_recvmsg().
If no control message structure is setup, kTLS layer will read and
process TLS data record types. As soon as it encounters a TLS control
message, it would return an error. At that point, NFS can setup a kvec
backed control buffer and read in the control message such as a TLS
alert. Scott found that a msg iterator can advance the kvec pointer
as a part of the copy process thus we need to revert the iterator
before calling into the tls_alert_recv.
md_spares_need_change in md_start_sync will call rdev_addable which
protected by rcu_read_lock/rcu_read_unlock. This rcu context will help
protect rdev won't be released, but rdev->mddev will be set to NULL
before we call synchronize_rcu in md_kick_rdev_from_array. Fix this by
using READ_ONCE and check does rdev->mddev still alive.
Fixes: bc08041b32ab ("md: suspend array in md_start_sync() if array need reconfiguration") Fixes: 570b9147deb6 ("md: use RCU lock to protect traversal in md_spares_need_change()") Signed-off-by: Yang Erkun <yangerkun@huawei.com> Link: https://lore.kernel.org/linux-raid/20250731114530.776670-1-yangerkun@huawei.com Signed-off-by: Yu Kuai <yukuai3@huawei.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Syzbot reported a WARNING in taprio_get_start_time().
When link speed is 470,589 or greater, q->picos_per_byte becomes too
small, causing length_to_duration(q, ETH_ZLEN) to return zero.
This zero value leads to validation failures in fill_sched_entry() and
parse_taprio_schedule(), allowing arbitrary values to be assigned to
entry->interval and cycle_time. As a result, sched->cycle can become zero.
Since SPEED_800000 is the largest defined speed in
include/uapi/linux/ethtool.h, this issue can occur in realistic scenarios.
To ensure length_to_duration() returns a non-zero value for minimum-sized
Ethernet frames (ETH_ZLEN = 60), picos_per_byte must be at least 17
(60 * 17 > PSEC_PER_NSEC which is 1000).
This patch enforces a minimum value of 17 for picos_per_byte when the
calculated value would be lower, and adds a warning message to inform
users that scheduling accuracy may be affected at very high link speeds.
When sending a packet with virtio_net_hdr to tun device, if the gso_type
in virtio_net_hdr is SKB_GSO_UDP and the gso_size is less than udphdr
size, below crash may happen.
To trigger gso segment in udp_queue_rcv_skb(), we should also set option
UDP_ENCAP_ESPINUDP to enable udp_sk(sk)->encap_rcv. When the encap_rcv
hook return 1 in udp_queue_rcv_one_skb(), udp_csum_pull_header() will try
to pull udphdr, but the skb size has been segmented to gso size, which
leads to this crash.
Previous commit cf329aa42b66 ("udp: cope with UDP GRO packet misdirection")
introduces segmentation in UDP receive path only for GRO, which was never
intended to be used for UFO, so drop UFO packets in udp_rcv_segment().
When the parent clock is a gated clock which has multiple parents, the
clock provider (clk-scmi typically) might return a rate of 0 since there
is not one of those particular parent clocks that should be chosen for
returning a rate. Prior to ee975351cf0c ("net: mdio: mdio-bcm-unimac:
Manage clock around I/O accesses"), we would not always be passing a
clock reference depending upon how mdio-bcm-unimac was instantiated. In
that case, we would take the fallback path where the rate is hard coded
to 250MHz.
Make sure that we still fallback to using a fixed rate for the divider
calculation, otherwise we simply ignore the desired MDIO bus clock
frequency which can prevent us from interfacing with Ethernet PHYs
properly.
Fixes: ee975351cf0c ("net: mdio: mdio-bcm-unimac: Manage clock around I/O accesses") Signed-off-by: Florian Fainelli <florian.fainelli@broadcom.com> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Reviewed-by: Simon Horman <horms@kernel.org> Link: https://patch.msgid.link/20250730202533.3463529-1-florian.fainelli@broadcom.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
When gso_segs is left at 0, a number of assumptions will end up being
incorrect throughout the stack.
For example, in the GRO-path, we set NAPI_GRO_CB()->count to gso_segs.
So, if a non-LRO'ed packet followed by an LRO'ed packet is being
processed in GRO, the first one will have NAPI_GRO_CB()->count set to 1 and
the next one to 0 (in dev_gro_receive()).
Since commit 531d0d32de3e
("net/mlx5: Correctly set gso_size when LRO is used")
these packets will get merged (as their gso_size now matches).
So, we end up in gro_complete() with NAPI_GRO_CB()->count == 1 and thus
don't call inet_gro_complete(). Meaning, checksum-validation in
tcp_checksum_complete() will fail with a "hw csum failure".
Even before the above mentioned commit, incorrect gso_segs means that other
things like TCP's accounting of incoming packets (tp->segs_in,
data_segs_in, rcv_ooopack) will be incorrect. Which means that if one
does bytes_received/data_segs_in, the result will be bigger than the
MTU.
Fix this by initializing gso_segs correctly when LRO is used.
Fixes: e586b3b0baee ("net/mlx5: Ethernet Datapath files") Reported-by: Gal Pressman <gal@nvidia.com> Closes: https://lore.kernel.org/netdev/6583783f-f0fb-4fb1-a415-feec8155bc69@nvidia.com/ Signed-off-by: Christoph Paasch <cpaasch@openai.com> Reviewed-by: Gal Pressman <gal@nvidia.com> Reviewed-by: Eric Dumazet <edumazet@google.com> Link: https://patch.msgid.link/20250729-mlx5_gso_segs-v1-1-b48c480c1c12@openai.com Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
In order to avoid any possible race we need to hold the ppe_lock
spinlock accessing the hw PPE table. airoha_ppe_foe_get_entry routine is
always executed holding ppe_lock except in airoha_ppe_debugfs_foe_show
routine. Fix the problem introducing airoha_ppe_foe_get_entry_locked
routine.
Fixes: 3fe15c640f380 ("net: airoha: Introduce PPE debugfs support") Reviewed-by: Dawid Osuchowski <dawid.osuchowski@linux.intel.com> Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org> Link: https://patch.msgid.link/20250731-airoha_ppe_foe_get_entry_locked-v2-1-50efbd8c0fd6@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The software node does not specify a count of property entries, so the
array must be null-terminated.
When unterminated, this can lead to a fault in the downstream cs35l56
amplifier driver, because the node parse walks off the end of the
array into unknown memory.
Fixes: 0ca645ab5b15 ("spi: cs42l43: Add speaker id support to the bridge configuration") Closes: https://bugzilla.kernel.org/show_bug.cgi?id=220371 Signed-off-by: Simon Trimmer <simont@opensource.cirrus.com> Link: https://patch.msgid.link/20250731160109.1547131-1-simont@opensource.cirrus.com Signed-off-by: Mark Brown <broonie@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
What: /sys/block/<disk>/queue/discard_granularity
[...]
A discard_granularity of 0 means that the device does not support
discard functionality.
but this got broken when sorting out the block limits updates. Fix this
by setting the discard_granularity limit to zero when the combined
max_discard_sectors is zero.
Fixes: 3c407dc723bb ("block: default the discard granularity to sector size") Signed-off-by: Christoph Hellwig <hch@lst.de> Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com> Link: https://lore.kernel.org/r/20250731152228.873923-1-hch@lst.de Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Sasha Levin <sashal@kernel.org>
Upstream commit 53889bcaf536 ("block: make __get_task_ioprio() easier to
read") changes the IO priority returned to the caller if no IO context
is defined for the task. Prior to this commit, the returned IO priority
was determined by task_nice_ioclass() and task_nice_ioprio(). Now it is
always IOPRIO_DEFAULT, which translates to IOPRIO_CLASS_NONE with priority
0. However, task_nice_ioclass() returns IOPRIO_CLASS_IDLE, IOPRIO_CLASS_RT,
or IOPRIO_CLASS_BE depending on the task scheduling policy, and
task_nice_ioprio() returns a value determined by task_nice(). This causes
regressions in test code checking the IO priority and class of IO
operations on tasks with no IO context.
Fix the problem by returning the IO priority calculated from
task_nice_ioclass() and task_nice_ioprio() if no IO context is defined
to match earlier behavior.
Fixes: 53889bcaf536 ("block: make __get_task_ioprio() easier to read") Cc: Jens Axboe <axboe@kernel.dk> Cc: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Guenter Roeck <linux@roeck-us.net> Reviewed-by: Yu Kuai <yukuai3@huawei.com> Reviewed-by: Damien Le Moal <dlemoal@kernel.org> Link: https://lore.kernel.org/r/20250731044953.1852690-1-linux@roeck-us.net Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Sasha Levin <sashal@kernel.org>
Module (SFP) eeprom GET has a lot of input params, they are all
mistakenly listed as output in the spec. Looks like kernel doesn't
output them at all. Correct what are the inputs and what the outputs.
Reported-by: Duo Yi <duo@meta.com> Fixes: a353318ebf24 ("tools: ynl: populate most of the ethtool spec") Acked-by: Stanislav Fomichev <sdf@fomichev.me> Link: https://patch.msgid.link/20250730172137.1322351-1-kuba@kernel.org Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The value of high_memory variable is set by set_high_memory() function
to a value returned by memblock_end_of_DRAM(). The latter function
returns by default the upper bound of the last online memory block,
not the upper bound of the directly mapped memory region. As result,
in case the end of memory happens to be offline, high_memory variable
is set to a value that is short on the last offline memory blocks size:
RANGE SIZE STATE REMOVABLE BLOCK
0x0000000000000000-0x000000ffffffffff 1T online yes 0-511
0x0000010000000000-0x0000011fffffffff 128G offline 512-575
Memory block size: 2G
Total online memory: 1T
Total offline memory: 128G
In the past the value of high_memory was derived from max_low_pfn,
which in turn was derived from the identity_size. Since identity_size
accommodates the whole memory size - including tailing offline blocks,
the offlined blocks did not impose any problem. But since commit e120d1bc12da ("arch, mm: set high_memory in free_area_init()") the
value of high_memory is derived from the last memblock online region,
and that is where the problem comes from.
The value of high_memory is used by several drivers and by external
tools (e.g. crash tool aborts while loading a dump).
Similarily to ARM, use the override path provided by set_high_memory()
function and set the value of high_memory at the end of the identity
mapping early. That forces set_high_memory() to leave in high_memory
the correct value, even when the end of available memory is offline.
Fixes: e120d1bc12da ("arch, mm: set high_memory in free_area_init()") Tested-by: Mikhail Zaslonko <zaslonko@linux.ibm.com> Reviewed-by: Heiko Carstens <hca@linux.ibm.com> Reviewed-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com> Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The SLCF bit ("stateless command filtering") introduced with
CEX8 cards was because of the function mask's default value
suppressed when user space read the ap function for an AP
card or queue. Unmask this bit so that user space applications
like lszcrypt can evaluate and list this feature.
Fixes: d4c53ae8e494 ("s390/ap: store TAPQ hwinfo in struct ap_card") Signed-off-by: Harald Freudenberger <freude@linux.ibm.com> Reviewed-by: Holger Dengler <dengler@linux.ibm.com> Signed-off-by: Heiko Carstens <hca@linux.ibm.com> Signed-off-by: Alexander Gordeev <agordeev@linux.ibm.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
During nvme target initialization discovery subsystem is initialized
before "nvmet" debugfs directory is created. This results in discovery
subsystem debugfs directory to be created in debugfs root directory.
In other words, the codepath above is exeucted before nvmet_debugfs is
created. We get /sys/kernel/debug/nqn.2014-08.org.nvmexpress.discovery
instead of /sys/kernel/debug/nvmet/nqn.2014-08.org.nvmexpress.discovery.
Move nvmet_init_discovery() call after nvmet_init_debugfs() to fix it.
Fixes: 649fd41420a8 ("nvmet: add debugfs support") Signed-off-by: Mohamed Khalfella <mkhalfella@purestorage.com> Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com> Reviewed-by: Hannes Reinecke <hare@kernel.org> Reviewed-by: Daniel Wagner <dwagner@suse.de> Signed-off-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit bafbdd527d56 ("phylib: Add device reset GPIO support") removed
devm_gpiod_get_optional() in favor of the non-devres managed
fwnode_get_named_gpiod(). When it was kind-of reverted by commit 40ba6a12a548 ("net: mdio: switch to using gpiod_get_optional()"), the devm
functionality was not reinstated. Nor was the GPIO unclaimed on device
remove. This leads to the GPIO being claimed indefinitely, even when the
device and/or the driver gets removed.
Fixes: bafbdd527d56 ("phylib: Add device reset GPIO support") Fixes: 40ba6a12a548 ("net: mdio: switch to using gpiod_get_optional()") Cc: Csaba Buday <buday.csaba@prolan.hu> Signed-off-by: Bence Csókás <csokas.bence@prolan.hu> Reviewed-by: Andrew Lunn <andrew@lunn.ch> Link: https://patch.msgid.link/20250728153455.47190-2-csokas.bence@prolan.hu Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
According to the 1588 standard, it is possible to use both unicast and
multicast frames to send the PTP information. It was noticed that if the
frames were unicast they were not processed by the analyzer meaning that
they were not timestamped. Therefore fix this to match also these
unicast frames.
Paolo spotted hangs in NIPA running driver tests against virtio.
The tests hang in virtnet_close() -> virtnet_napi_tx_disable().
The problem is only reproducible if running multiple of our tests
in sequence (I used TEST_PROGS="xdp.py ping.py netcons_basic.sh \
netpoll_basic.py stats.py"). Initial suspicion was that this is
a simple case of double-disable of NAPI, but instrumenting the
code reveals:
Deadlocked on NAPI ffff888007cd82c0 (virtnet_poll_tx):
state: 0x37, disabled: false, owner: 0, listed: false, weight: 64
The NAPI was not in fact disabled, owner is 0 (rather than -1),
so the NAPI "thinks" it's scheduled for CPU 0 but it's not listed
(!list_empty(&n->poll_list) => false). It seems odd that normal NAPI
processing would wedge itself like this.
Better suspicion is that netpoll gets enabled while NAPI is polling,
and also grabs the NAPI instance. This confuses napi_complete_done():
[netpoll] [normal NAPI]
napi_poll()
have = netpoll_poll_lock()
rcu_access_pointer(dev->npinfo)
return NULL # no netpoll
__napi_poll()
->poll(->weight)
poll_napi()
cmpxchg(->poll_owner, -1, cpu)
poll_one_napi()
set_bit(NAPI_STATE_NPSVC, ->state)
napi_complete_done()
if (NAPIF_STATE_NPSVC)
return false
# exit without clearing SCHED
This feels very unlikely, but perhaps virtio has some interactions
with the hypervisor in the NAPI ->poll that makes the race window
larger?
Best I could to to prove the theory was to add and trigger this
warning in napi_poll (just before netpoll_poll_unlock()):
If this warning hits the next virtio_close() will hang.
This patch survived 30 test iterations without a hang (without it
the longest clean run was around 10). Credit for triggering this
goes to Breno's recent netconsole tests.
Flags passed in for splice() syscall should not end up in
skb_recv_datagram(). As SPLICE_F_NONBLOCK == MSG_PEEK, kernel gets
confused: skb isn't unlinked from a receive queue, while strp_msg::offset
and strp_msg::full_len are updated.
Unbreak the logic a bit more by mapping both O_NONBLOCK and
SPLICE_F_NONBLOCK to MSG_DONTWAIT. This way we align with man splice(2) in
regard to errno EAGAIN:
SPLICE_F_NONBLOCK was specified in flags or one of the file descriptors
had been marked as nonblocking (O_NONBLOCK), and the operation would
block.
Fixes: 5121197ecc5d ("kcm: close race conditions on sk_receive_queue") Fixes: 91687355b927 ("kcm: Splice support") Signed-off-by: Michal Luczaj <mhal@rbox.co> Link: https://patch.msgid.link/20250725-kcm-splice-v1-1-9a725ad2ee71@rbox.co Signed-off-by: Jakub Kicinski <kuba@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit a1fd37f97808 ("md: Don't wait for MD_RECOVERY_NEEDED for
HOT_REMOVE_DISK ioctl") introduced a regression in the md_cluster
module. (Failed cases 02r1_Manage_re-add & 02r10_Manage_re-add)
Consider a 2-node cluster:
- node1 set faulty & remove command on a disk.
- node2 must correctly update the array metadata.
Before a1fd37f97808, on node1, the delay between msg:METADATA_UPDATED
(triggered by faulty) and msg:REMOVE was sufficient for node2 to
reload the disk info (written by node1).
After a1fd37f97808, node1 no longer waits between faulty and remove,
causing it to send msg:REMOVE while node2 is still reloading disk info.
This often results in node2 failing to remove the faulty disk.
== how to trigger ==
set up a 2-node cluster (node1 & node2) with disks vdc & vdd.
check array status on both nodes with "mdadm -D /dev/md0".
node1 output:
Number Major Minor RaidDevice State
- 0 0 0 removed
1 254 48 1 active sync /dev/vdd
node2 output:
Number Major Minor RaidDevice State
- 0 0 0 removed
1 254 48 1 active sync /dev/vdd
0 254 32 - faulty /dev/vdc
Fixes: a1fd37f97808 ("md: Don't wait for MD_RECOVERY_NEEDED for HOT_REMOVE_DISK ioctl") Signed-off-by: Heming Zhao <heming.zhao@suse.com> Reviewed-by: Su Yue <glass.su@suse.com> Link: https://lore.kernel.org/linux-raid/20250728042145.9989-1-heming.zhao@suse.com Signed-off-by: Yu Kuai <yukuai3@huawei.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit fb52f3226cab ("ARM: s3c/gpio: use new line value setter
callbacks") correctly changed the assignment of the callback but missed
the check one liner higher. Change it now too to using the recommended
callback as the legacy one is going away soon.
Fixes: fb52f3226cab ("ARM: s3c/gpio: use new line value setter callbacks") Signed-off-by: Bartosz Golaszewski <bartosz.golaszewski@linaro.org> Signed-off-by: Arnd Bergmann <arnd@arndb.de> Signed-off-by: Sasha Levin <sashal@kernel.org>
If the NFS client is doing writeback from a workqueue context, avoid using
__GFP_NORETRY for allocations if the task has set PF_MEMALLOC_NOIO or
PF_MEMALLOC_NOFS. The combination of these flags makes memory allocation
failures much more likely.
We've seen those allocation failures show up when the loopback driver is
doing writeback from a workqueue to a file on NFS, where memory allocation
failure results in errors or corruption within the loopback device's
filesystem.
Currently, when the server supports NFS4.1 security labels then
security.selinux label in included twice. Instead, only add it
when the server doesn't possess security label support.
When an applications get killed (SIGTERM/SIGINT) while pNFS client performs a connection
to DS, client ends in an infinite loop of connect-disconnect. This
source of the issue, it that flexfilelayoutdev#nfs4_ff_layout_prepare_ds gets an error
on nfs4_pnfs_ds_connect with status ERESTARTSYS, which is set by rpc_signal_task, but
the error is treated as transient, thus retried.
The issue is reproducible with Ctrl+C the following script(there should be ~1000 files in
a directory, client should must not have any connections to DSes):
```
echo 3 > /proc/sys/vm/drop_caches
for i in *
do
head -1 $i
done
```
The change aims to propagate the nfs4_ff_layout_prepare_ds error state
to the caller that can decide whatever this is a retryable error or not.
Signed-off-by: Tigran Mkrtchyan <tigran.mkrtchyan@desy.de> Link: https://lore.kernel.org/r/20250627071751.189663-1-tigran.mkrtchyan@desy.de Fixes: 260f32adb88d ("pNFS/flexfiles: Check the result of nfs4_pnfs_ds_connect") Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The existing PowerNV hotplug code did not handle surprise plug events
correctly, leading to a complete failure of the hotplug system after device
removal and a required reboot to detect new devices.
This comes down to two issues:
1) When a device is surprise removed, often the bridge upstream
port will cause a PE freeze on the PHB. If this freeze is not
cleared, the MSI interrupts from the bridge hotplug notification
logic will not be received by the kernel, stalling all plug events
on all slots associated with the PE.
2) When a device is removed from a slot, regardless of surprise or
programmatic removal, the associated PHB/PE ls left frozen.
If this freeze is not cleared via a fundamental reset, skiboot
is unable to clear the freeze and cannot retrain / rescan the
slot. This also requires a reboot to clear the freeze and redetect
the device in the slot.
Issue the appropriate unfreeze and rescan commands on hotplug events,
and don't oops on hotplug if pci_bus_to_OF_node() returns NULL.
The Microsemi Switchtec PM8533 PFX 48xG3 [11f8:8533] PCIe switch system
was observed to incorrectly assert the Presence Detect Set bit in its
capabilities when tested on a Raptor Computing Systems Blackbird system,
resulting in the hot insert path never attempting a rescan of the bus
and any downstream devices not being re-detected.
Work around this by additionally checking whether the PCIe data link is
active or not when performing presence detection on downstream switches'
ports, similar to the pciehp_hpc.c driver.
When the root of a nested PCIe bridge configuration is unplugged, the
pnv_php driver leaked the allocated IRQ resources for the child bridges'
hotplug event notifications, resulting in a panic.
Fix this by walking all child buses and deallocating all its IRQ resources
before calling pci_hp_remove_devices().
Also modify the lifetime of the workqueue at struct pnv_php_slot::wq so
that it is only destroyed in pnv_php_free_slot(), instead of
pnv_php_disable_irq(). This is required since pnv_php_disable_irq() will
now be called by workers triggered by hot unplug interrupts, so the
workqueue needs to stay allocated.
The abridged kernel panic that occurs without this patch is as follows:
Remove comment for reorder_work which no longer exists.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> Fixes: 71203f68c774 ("padata: Fix pd UAF once and for all") Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Sasha Levin <sashal@kernel.org>
This was missed during the initial implementation. The VFIO PCI encodes
the vf_token inside the device name when opening the device from the group
FD, something like:
This is used to control access to a VF unless there is co-ordination with
the owner of the PF.
Since we no longer have a device name in the cdev path, pass the token
directly through VFIO_DEVICE_BIND_IOMMUFD using an optional field
indicated by VFIO_DEVICE_BIND_FLAG_TOKEN.
Fixes: 5fcc26969a16 ("vfio: Add VFIO_DEVICE_BIND_IOMMUFD") Tested-by: Shameer Kolothum <shameerali.kolothum.thodi@huawei.com> Reviewed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Jason Gunthorpe <jgg@nvidia.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Link: https://lore.kernel.org/r/0-v3-bdd8716e85fe+3978a-vfio_token_jgg@nvidia.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Commit aa3998dbeb3a ("ata: libata-scsi: Disable scsi device
manage_system_start_stop") enabled libata EH to manage device power mode
trasitions for system suspend/resume and removed the flag from
ata_scsi_dev_config. However, since the sd_shutdown() function still
relies on the manage_system_start_stop flag, a spin-down command is not
issued to the disk with command "echo 1 > /sys/block/sdb/device/delete"
sd_shutdown() can be called for both system/runtime start stop
operations, so utilize the manage_run_time_start_stop flag set in the
ata_scsi_dev_config and issue a spin-down command during disk removal
when the system is running. This is in addition to when the system is
powering off and manage_shutdown flag is set. The
manage_system_start_stop flag will still be used for drivers that still
set the flag.
Fixes: aa3998dbeb3a ("ata: libata-scsi: Disable scsi device manage_system_start_stop") Signed-off-by: Salomon Dushimirimana <salomondush@google.com> Link: https://lore.kernel.org/r/20250724214520.112927-1-salomondush@google.com Tested-by: Damien Le Moal <dlemoal@kernel.org> Reviewed-by: Damien Le Moal <dlemoal@kernel.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
If the h8 exit fails during runtime resume process, the runtime thread
enters runtime suspend immediately and the error handler operates at the
same time. It becomes stuck and cannot be recovered through the error
handler. To fix this, use link recovery instead of the error handler.
Fixes: 4db7a2360597 ("scsi: ufs: Fix concurrency of error handler and other error recovery paths") Signed-off-by: Seunghui Lee <sh043.lee@samsung.com> Link: https://lore.kernel.org/r/20250717081213.6811-1-sh043.lee@samsung.com Reviewed-by: Bean Huo <beanhuo@micron.com> Acked-by: Bart Van Assche <bvanassche@acm.org> Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
The invocation of iscsi_put_conn() in iscsi_iter_destory_conn_fn() is
used to free the initial reference counter of iscsi_cls_conn. For
non-qla4xxx cases, the ->destroy_conn() callback (e.g.,
iscsi_conn_teardown) will call iscsi_remove_conn() and iscsi_put_conn()
to remove the connection from the children list of session and free the
connection at last. However for qla4xxx, it is not the case. The
->destroy_conn() callback of qla4xxx will keep the connection in the
session conn_list and doesn't use iscsi_put_conn() to free the initial
reference counter. Therefore, it seems necessary to keep the
iscsi_put_conn() in the iscsi_iter_destroy_conn_fn(), otherwise, there
will be memory leak problem.
In the below noted Fixes commit we introduced a reflck mutex to allow
better scaling between devices for open and close. The reflck was
based on the hot reset granularity, device level for root bus devices
which cannot support hot reset or bus/slot reset otherwise. Overlooked
in this were SR-IOV VFs, where there's also no bus reset option, but
the default for a non-root-bus, non-slot-based device is bus level
reflck granularity.
The reflck mutex has since become the dev_set mutex (via commit 2cd8b14aaa66 ("vfio/pci: Move to the device set infrastructure")) and
is our defacto serialization for various operations and ioctls. It
still seems to be the case though that sets of vfio-pci devices really
only need serialization relative to hot resets affecting the entire
set, which is not relevant to SR-IOV VFs. As described in the Closes
link below, this serialization contributes to startup latency when
multiple VFs sharing the same "bus" are opened concurrently.
Mark the device itself as the basis of the dev_set for SR-IOV VFs.
Reported-by: Aaron Lewis <aaronlewis@google.com> Closes: https://lore.kernel.org/all/20250626180424.632628-1-aaronlewis@google.com Tested-by: Aaron Lewis <aaronlewis@google.com> Fixes: e309df5b0c9e ("vfio/pci: Parallelize device open and release") Reviewed-by: Yi Liu <yi.l.liu@intel.com> Reviewed-by: Kevin Tian <kevin.tian@intel.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Link: https://lore.kernel.org/r/20250626225623.1180952-1-alex.williamson@redhat.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
When vfio_df_close() is called with open_count=0, it triggers a warning in
vfio_assert_device_open() but still decrements open_count to -1. This allows
a subsequent open to incorrectly pass the open_count == 0 check, leading to
unintended behavior, such as setting df->access_granted = true.
For example, running an IOMMUFD compat no-IOMMU device with VFIO tests
(https://github.com/awilliam/tests/blob/master/vfio-noiommu-pci-device-open.c)
results in a warning and a failed VFIO_GROUP_GET_DEVICE_FD ioctl on the first
run, but the second run succeeds incorrectly.
Add checks to avoid decrementing open_count below zero.
Fixes: 05f37e1c03b6 ("vfio: Pass struct vfio_device_file * to vfio_device_open/close()") Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Reviewed-by: Yi Liu <yi.l.liu@intel.com> Signed-off-by: Jacob Pan <jacob.pan@linux.microsoft.com> Link: https://lore.kernel.org/r/20250618234618.1910456-2-jacob.pan@linux.microsoft.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
For devices with no-iommu enabled in IOMMUFD VFIO compat mode, the group open
path skips vfio_df_open(), leaving open_count at 0. This causes a warning in
vfio_assert_device_open(device) when vfio_df_close() is called during group
close.
The correct behavior is to skip only the IOMMUFD bind in the device open path
for no-iommu devices. Commit 6086efe73498 omitted vfio_df_open(), which was
too broad. This patch restores the previous behavior, ensuring
the vfio_df_open is called in the group open path.
Fixes: 6086efe73498 ("vfio-iommufd: Move noiommu compat validation out of vfio_iommufd_bind()") Suggested-by: Alex Williamson <alex.williamson@redhat.com> Suggested-by: Jason Gunthorpe <jgg@nvidia.com> Signed-off-by: Jacob Pan <jacob.pan@linux.microsoft.com> Reviewed-by: Jason Gunthorpe <jgg@nvidia.com> Link: https://lore.kernel.org/r/20250618234618.1910456-1-jacob.pan@linux.microsoft.com Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Sasha Levin <sashal@kernel.org>
Test: androbench by default setting, use 64GB sdcard.
the random write speed:
without this patch 3.5MB/s
with this patch 7MB/s
After patch "11a347fb6cef", the random write speed decreased significantly.
the .write_iter() interface had been modified, and check the differences
with generic_file_write_iter(), when calling generic_write_sync() and
exfat_file_write_iter() to call vfs_fsync_range(), the fdatasync flag is
wrong, and make not use the fdatasync mode, and make random write speed
decreased. So use generic_write_sync() instead of vfs_fsync_range().
Fixes: 11a347fb6cef ("exfat: change to get file size from DataLength") Signed-off-by: Zhengxu Zhang <zhengxu.zhang@unisoc.com> Acked-by: Yuezhang Mo <Yuezhang.Mo@sony.com> Signed-off-by: Namjae Jeon <linkinjeon@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
The root cause of we run out-of-space is: in f2fs_map_blocks(), f2fs may
trigger foreground gc only if it allocates any physical block, it will be
a little bit later when there is multiple threads writing data w/
aio/dio/bufio method in parallel, since we always use OPU in lfs mode, so
f2fs_map_blocks() does block allocations aggressively.
In order to fix this issue, let's give a chance to trigger foreground
gc in prior to block allocation in f2fs_map_blocks().
In lfs mode, dirty data needs OPU, we'd better calculate lower_p and
upper_p w/ them during has_not_enough_free_secs(), otherwise we may
encounter out-of-space issue due to we missed to reclaim enough
free section w/ foreground gc.
When testing F2FS with xfstests using UFS backed virtual disks the
kernel complains sometimes that f2fs_release_decomp_mem() calls
vm_unmap_ram() from an invalid context. Example trace from
f2fs/007 test:
This patch modifies in_task() check inside f2fs_read_end_io() to also
check if interrupts are disabled. This ensures that pages are unmapped
asynchronously in an interrupt handler.
Fixes: bff139b49d9f ("f2fs: handle decompress only post processing in softirq") Signed-off-by: Jan Prusakowski <jprusakowski@google.com> Reviewed-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
If device path length equals to MAX_PATH_LEN, sbi->devs.path[] may
not end up w/ null character due to path array is fully filled, So
accidently, fields locate after path[] may be treated as part of
device path, result in parsing wrong device path.
The buggy address belongs to the object at ffff88812d961f20
which belongs to the cache f2fs_inode_cache of size 1200
The buggy address is located 856 bytes inside of
1200-byte region [ffff88812d961f20, ffff88812d9623d0)
This bug can be reproduced w/ the reproducer [2], once we enable
CONFIG_F2FS_CHECK_FS config, the reproducer will trigger panic as below,
so the direct reason of this bug is the same as the one below patch [3]
fixed.
The root cause is: in the fuzzed image, dnode #8 belongs to inode #7,
after inode #7 eviction, dnode #8 was dropped.
However there is dirent that has ino #8, so, once we unlink file3, in
f2fs_evict_inode(), both f2fs_truncate() and f2fs_update_inode_page()
will fail due to we can not load node #8, result in we missed to call
f2fs_inode_synced() to clear inode dirty status.
Let's fix this by calling f2fs_inode_synced() in error path of
f2fs_evict_inode().
PS: As I verified, the reproducer [2] can trigger this bug in v6.1.129,
but it failed in v6.16-rc4, this is because the testcase will stop due to
other corruption has been detected by f2fs:
F2FS-fs (loop0): inconsistent node block, node_type:2, nid:8, node_footer[nid:8,ino:8,ofs:0,cpver:5013063228981249506,blkaddr:15366]
F2FS-fs (loop0): f2fs_lookup: inode (ino=9) has zero i_nlink
Fixes: 0f18b462b2e5 ("f2fs: flush inode metadata when checkpoint is doing") Closes: https://syzkaller.appspot.com/x/report.txt?x=13448368580000 Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
==================================================================
BUG: KASAN: use-after-free in __list_del_entry_valid+0xa6/0x130 lib/list_debug.c:62
Read of size 8 at addr ffff888100567dc8 by task kworker/u4:0/8
The buggy address belongs to the object at ffff888100567a10
which belongs to the cache f2fs_inode_cache of size 1360
The buggy address is located 952 bytes inside of
1360-byte region [ffff888100567a10, ffff888100567f60)
The root cause is w/ a fuzzed image, f2fs may missed to clear FI_DIRTY_INODE
flag for target inode, after f2fs_evict_inode(), the inode is still linked in
sbi->inode_list[DIRTY_META] global list, once it triggers checkpoint,
f2fs_sync_inode_meta() may access the released inode.
In f2fs_evict_inode(), let's always call f2fs_inode_synced() to clear
FI_DIRTY_INODE flag and drop inode from global dirty list to avoid this
UAF issue.
Fixes: 0f18b462b2e5 ("f2fs: flush inode metadata when checkpoint is doing") Closes: https://syzkaller.appspot.com/bug?extid=849174b2efaf0d8be6ba Signed-off-by: Chao Yu <chao@kernel.org> Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org> Signed-off-by: Sasha Levin <sashal@kernel.org>
projects / thirdparty / kernel / stable.git / log
