Bringing a linked netdevsim device down and then up causes communication
failure because both interfaces lack carrier. Basically a ifdown/ifup on
the interface make the link broken.
Commit 3762ec05a9 ("netdevsim: add NAPI support") added supported
for NAPI, calling netif_carrier_off() in nsim_stop(). This patch
re-enables the carrier symmetrically on nsim_open(), in case the device
is linked and the peer is up.
Signed-off-by: Breno Leitao <leitao@debian.org>
Fixes: 3762ec05a9 ("netdevsim: add NAPI support")
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Link: https://patch.msgid.link/20251014-netdevsim_fix-v2-1-53b40590dae1@debian.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Sabrina Dubroca says:
====================
tls: misc bugfixes
Jann Horn reported multiple bugs in kTLS. This series addresses them,
and adds some corresponding selftests for those that are reproducible
(and without failure injection).
====================
Link: https://patch.msgid.link/cover.1760432043.git.sd@queasysnail.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
With async crypto, we rely on tx_work to actually transmit records
once encryption completes. But while send() is running, both the
tx_lock and socket lock are held, so tx_work_handler cannot process
the queue of encrypted records, and simply reschedules itself. During
a large send(), this could last a long time, and use a lot of memory.
Transmit any pending encrypted records before restarting the main
loop of tls_sw_sendmsg_locked.
Fixes: a42055e8d2 ("net/tls: Add support for async encryption of records for performance")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/8396631478f70454b44afb98352237d33f48d34d.1760432043.git.sd@queasysnail.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Async decryption calls tls_strp_msg_hold to create a clone of the
input skb to hold references to the memory it uses. If we fail to
allocate that clone, proceeding with async decryption can lead to
various issues (UAF on the skb, writing into userspace memory after
the recv() call has returned).
In this case, wait for all pending decryption requests.
Fixes: 84c61fe1a7 ("tls: rx: do not use the standard strparser")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/b9fe61dcc07dab15da9b35cf4c7d86382a98caf2.1760432043.git.sd@queasysnail.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
When userspace wants to send a non-DATA record (via the
TLS_SET_RECORD_TYPE cmsg), we need to send any pending data from a
previous MSG_MORE send() as a separate DATA record. If that DATA record
is encrypted asynchronously, tls_handle_open_record will return
-EINPROGRESS. This is currently treated as an error by
tls_process_cmsg, and it will skip setting record_type to the correct
value, but the caller (tls_sw_sendmsg_locked) handles that return
value correctly and proceeds with sending the new message with an
incorrect record_type (DATA instead of whatever was requested in the
cmsg).
Always set record_type before handling the open record. If
tls_handle_open_record returns an error, record_type will be
ignored. If it succeeds, whether with synchronous crypto (returning 0)
or asynchronous (returning -EINPROGRESS), the caller will proceed
correctly.
Fixes: a42055e8d2 ("net/tls: Add support for async encryption of records for performance")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/0457252e578a10a94e40c72ba6288b3a64f31662.1760432043.git.sd@queasysnail.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
If we hit an error during the main loop of tls_sw_sendmsg_locked (eg
failed allocation), we jump to send_end and immediately
return. Previous iterations may have queued async encryption requests
that are still pending. We should wait for those before returning, as
we could otherwise be reading from memory that userspace believes
we're not using anymore, which would be a sort of use-after-free.
This is similar to what tls_sw_recvmsg already does: failures during
the main loop jump to the "wait for async" code, not straight to the
unlock/return.
Fixes: a42055e8d2 ("net/tls: Add support for async encryption of records for performance")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/c793efe9673b87f808d84fdefc0f732217030c52.1760432043.git.sd@queasysnail.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
During tls_sw_sendmsg_locked, we pre-allocate the encrypted message
for the size we're expecting to send during the current iteration, but
we may end up sending less, for example when splicing: if we're
getting the data from small fragments of memory, we may fill up all
the slots in the skmsg with less data than expected.
In this case, we need to trim the encrypted message to only the length
we actually need, to avoid pushing uninitialized bytes down the
underlying TCP socket.
Fixes: fe1e81d4f7 ("tls/sw: Support MSG_SPLICE_PAGES")
Reported-by: Jann Horn <jannh@google.com>
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Link: https://patch.msgid.link/66a0ae99c9efc15f88e9e56c1f58f902f442ce86.1760432043.git.sd@queasysnail.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
While associating, firmware needs the peer's receive capability
to calculate its own VHT transmit MCS. Currently, the host
sends this information via mcs->rx_mcs_set field, but firmware
actually reads it from mcs->tx_mcs_set field. This mismatch is
incorrect.
This issue has not caused failures so far because most peers
advertise identical TX and RX capabilities. Fix this by
assigning the value to tx_mcs_set as expected.
Additionally, the rate control mask is intended to limit our
transmit MCS, so it should also apply to the peer's receive
capability. Update the logic accordingly.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.0.1-00029-QCAHKSWPL_SILICONZ-1
Fixes: d889913205 ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Signed-off-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Signed-off-by: Pradeep Kumar Chitrapu <quic_pradeepc@quicinc.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Link: https://patch.msgid.link/20251009211656.2386085-2-quic_pradeepc@quicinc.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Currently, in ext_rx_stats, the NSS value is taken directly from
the firmware, which results in incorrect mapping:
4x4, 3x3, 2x2, 1x1 SS are incorrectly updated as 3x3, 2x2,
1x1, 0x0 SS respectively.
Fix the issue by incrementing the NSS value by 1 while updating
the PPDU info to ensure accurate spatial stream statistics.
Remove the redundant +1 increment in the radiotap header when monitor
mode is enabled to prevent double counting.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.5-01651-QCAHKSWPL_SILICONZ-1
Signed-off-by: Thiraviyam Mariyappan <thiraviyam.mariyappan@oss.qualcomm.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20251007133332.1092178-1-thiraviyam.mariyappan@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Mac80211 schedules CSA finalize work twice during a channel switch: first
during the reserved switch phase and again during the finalize phase.
The beacon content is updated only during the second schedule, which occurs
after the reserved switch completes. However, the ath12k driver attempts to
bring up the VDEV during the channel switch callback
(ath12k_mac_update_vif_chan()), which leads to premature installation of
stale beacon templates before the updated content is available.
This premature VDEV bring-up causes outdated beacon information to be
broadcast, which can result in updated channel parameters during the
transition. In MBSSID scenarios, this behavior is particularly problematic
because the transmitting interface's beacon must be updated before
non-transmitting interfaces are brought up. Failing to do so can lead to
beacon mismatches across interfaces.
Introduce a is_csa_in_progress flag to defer VDEV_UP until CSA finalize is
complete. Set this flag during the channel switch callback when CSA is
active. In bss_info_changed(), check this flag and issue VDEV_UP only
after the beacon template has been updated.
Ensure that in MBSSID cases, the transmitting interface is brought up
first, followed by all non-transmitting interfaces. This ordering makes
sure correct beacon propagation and avoids stale beacon installation
during CSA transitions.
Additionally, move the call to ath12k_mac_update_peer_puncturing_width()
before VDEV bring-up during CSA handling. This ensures that the puncturing
bitmap and bandwidth settings are applied before the VDEV is brought up.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.3.1-00173-QCAHKSWPL_SILICONZ-1
Fixes: 8c6faa56bf ("wifi: ath12k: add MBSSID beacon support")
Signed-off-by: Aditya Kumar Singh <aditya.kumar.singh@oss.qualcomm.com>
Signed-off-by: Maharaja Kennadyrajan <maharaja.kennadyrajan@oss.qualcomm.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20250924134336.888-1-maharaja.kennadyrajan@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Currently, packets received on the REO exception ring from
unassociated peers are of MSDU buffer type, while the driver expects
link descriptor type packets. These packets are not parsed further due
to a return check on packet type in ath12k_hal_desc_reo_parse_err(),
but the associated skb is not freed. This may lead to kernel
crashes and buffer leaks.
Hence to fix, update the RX error handler to explicitly drop
MSDU buffer type packets received on the REO exception ring.
This prevents further processing of invalid packets and ensures
stability in the RX error handling path.
Tested-on: QCN9274 hw2.0 PCI WLAN.WBE.1.4.1-00199-QCAHKSWPL_SILICONZ-1
Fixes: d889913205 ("wifi: ath12k: driver for Qualcomm Wi-Fi 7 devices")
Signed-off-by: Sarika Sharma <sarika.sharma@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20250930091551.3305312-2-sarika.sharma@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Hardware target implements an address space larger than that PCI BAR can
map. In order to be able to access the whole target address space, the BAR
space is split into 4 segments, of which the last 3, called windows, can
be dynamically mapped to the desired area. This is achieved by updating
window register with appropriate window value. Currently each time when
accessing a register that beyond ATH11K_PCI_WINDOW_START, host calculates
the window value and caches it after window update, this way next time
when accessing a register falling in the same window, host knows that the
window is already good hence no additional update needed.
However this mechanism breaks after global reset is triggered in
ath11k_pci_soc_global_reset(), because with global reset hardware resets
window register hence the window is not properly mapped any more. Current
host does nothing about this, as a result a subsequent register access may
not work as expected if it falls in a window same as before.
Although there is no obvious issue seen now, better to fix it to avoid
future problem. The fix is done by restoring the window register after
global reset.
Tested-on: WCN6855 hw2.0 PCI WLAN.HSP.1.1-03125-QCAHSPSWPL_V1_V2_SILICONZ_LITE-3.6510.30
Fixes: d5c65159f2 ("ath11k: driver for Qualcomm IEEE 802.11ax devices")
Signed-off-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20251014-ath11k-reset-window-cache-v1-1-b85271b111dd@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Currently, ath10k has a recovery check logic. It will wait for the
last recovery to finish by wait_for_completion_timeout();
But in SDIO scenarios, the recovery function may be invoked from
interrupt context, where long blocking waits are undesirable and can
lead to system instability.
Additionally, Linux’s ordered workqueue processes one task at a time.
If a previous recovery is still queued or executing, new triggers are
ignored. This prevents accurate tracking of consecutive failures and
delays transition to the WEDGED state.
To address this, move the recovery check logic into a different
workqueue.
Tested-on: QCA6174 hw3.2 PCI WLAN.RM.4.4.1-00288-QCARMSWPZ-1
Tested-on: QCA6174 hw3.2 SDIO WLAN.RMH.4.4.1-00189
Fixes: c256a94d1b ("wifi: ath10k: shutdown driver when hardware is unreliable")
Signed-off-by: Kang Yang <kang.yang@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Link: https://patch.msgid.link/20251014110757.155-1-kang.yang@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
Pull vfs fixes from Christian Brauner:
- Handle inode number mismatches in nsfs file handles
- Update the comment to init_file()
- Add documentation link for EBADF in the rust file code
- Skip read lock assertion for read-only filesystems when using dax
- Don't leak disconnected dentries during umount
- Fix new coredump input pattern validation
- Handle ENOIOCTLCMD conversion in vfs_fileattr_{g,s}et() correctly
- Remove redundant IOCB_DIO_CALLER_COMP clearing in overlayfs
* tag 'vfs-6.18-rc2.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
ovl: remove redundant IOCB_DIO_CALLER_COMP clearing
fs: return EOPNOTSUPP from file_setattr/file_getattr syscalls
Revert "fs: make vfs_fileattr_[get|set] return -EOPNOTSUPP"
coredump: fix core_pattern input validation
vfs: Don't leak disconnected dentries on umount
dax: skip read lock assertion for read-only filesystems
rust: file: add intra-doc link for 'EBADF'
fs: update comment in init_file()
nsfs: handle inode number mismatches gracefully in file handles
When performing memory error injection on a THP (Transparent Huge Page)
mapped to userspace on an x86 server, the kernel panics with the following
trace. The expected behavior is to terminate the affected process instead
of panicking the kernel, as the x86 Machine Check code can recover from an
in-userspace #MC.
mce: [Hardware Error]: CPU 0: Machine Check Exception: f Bank 3: bd80000000070134
mce: [Hardware Error]: RIP 10:<ffffffff8372f8bc> {memchr_inv+0x4c/0xf0}
mce: [Hardware Error]: TSC afff7bbff88a ADDR 1d301b000 MISC 80 PPIN 1e741e77539027db
mce: [Hardware Error]: PROCESSOR 0:d06d0 TIME 1758093249 SOCKET 0 APIC 0 microcode 80000320
mce: [Hardware Error]: Run the above through 'mcelog --ascii'
mce: [Hardware Error]: Machine check: Data load in unrecoverable area of kernel
Kernel panic - not syncing: Fatal local machine check
The root cause of this panic is that handling a memory failure triggered
by an in-userspace #MC necessitates splitting the THP. The splitting
process employs a mechanism, implemented in
try_to_map_unused_to_zeropage(), which reads the pages in the THP to
identify zero-filled pages. However, reading the pages in the THP results
in a second in-kernel #MC, occurring before the initial memory_failure()
completes, ultimately leading to a kernel panic. See the kernel panic
call trace on the two #MCs.
First Machine Check occurs // [1]
memory_failure() // [2]
try_to_split_thp_page()
split_huge_page()
split_huge_page_to_list_to_order()
__folio_split() // [3]
remap_page()
remove_migration_ptes()
remove_migration_pte()
try_to_map_unused_to_zeropage() // [4]
memchr_inv() // [5]
Second Machine Check occurs // [6]
Kernel panic
[1] Triggered by accessing a hardware-poisoned THP in userspace, which is
typically recoverable by terminating the affected process.
[2] Call folio_set_has_hwpoisoned() before try_to_split_thp_page().
[3] Pass the RMP_USE_SHARED_ZEROPAGE remap flag to remap_page().
[4] Try to map the unused THP to zeropage.
[5] Re-access pages in the hw-poisoned THP in the kernel.
[6] Triggered in-kernel, leading to a panic kernel.
In Step[2], memory_failure() sets the poisoned flag on the page in the THP
by TestSetPageHWPoison() before calling try_to_split_thp_page().
As suggested by David Hildenbrand, fix this panic by not accessing to the
poisoned page in the THP during zeropage identification, while continuing
to scan unaffected pages in the THP for possible zeropage mapping. This
prevents a second in-kernel #MC that would cause kernel panic in Step[4].
Thanks to Andrew Zaborowski for his initial work on fixing this issue.
Link: https://lkml.kernel.org/r/20251015064926.1887643-1-qiuxu.zhuo@intel.com
Link: https://lkml.kernel.org/r/20251011075520.320862-1-qiuxu.zhuo@intel.com
Fixes: b1f202060a ("mm: remap unused subpages to shared zeropage when splitting isolated thp")
Signed-off-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Reported-by: Farrah Chen <farrah.chen@intel.com>
Suggested-by: David Hildenbrand <david@redhat.com>
Acked-by: David Hildenbrand <david@redhat.com>
Tested-by: Farrah Chen <farrah.chen@intel.com>
Tested-by: Qiuxu Zhuo <qiuxu.zhuo@intel.com>
Acked-by: Lance Yang <lance.yang@linux.dev>
Reviewed-by: Wei Yang <richard.weiyang@gmail.com>
Acked-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Cc: Barry Song <baohua@kernel.org>
Cc: Dev Jain <dev.jain@arm.com>
Cc: Jiaqi Yan <jiaqiyan@google.com>
Cc: Liam Howlett <liam.howlett@oracle.com>
Cc: Lorenzo Stoakes <lorenzo.stoakes@oracle.com>
Cc: "Luck, Tony" <tony.luck@intel.com>
Cc: Mariano Pache <npache@redhat.com>
Cc: Miaohe Lin <linmiaohe@huawei.com>
Cc: Naoya Horiguchi <nao.horiguchi@gmail.com>
Cc: Ryan Roberts <ryan.roberts@arm.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Commit 370645f41e ("dma-mapping: force bouncing if the kmalloc() size is
not cache-line-aligned") introduced DMA_BOUNCE_UNALIGNED_KMALLOC feature
and permitted architecture specific code configure kmalloc slabs with
sizes smaller than the value of dma_get_cache_alignment().
When that feature is enabled, the physical address of some small
kmalloc()-ed buffers might be not aligned to the CPU cachelines, thus not
really suitable for typical DMA. To properly handle that case a SWIOTLB
buffer bouncing is used, so no CPU cache corruption occurs. When that
happens, there is no point reporting a false-positive DMA-API warning that
the buffer is not properly aligned, as this is not a client driver fault.
[m.szyprowski@samsung.com: replace is_swiotlb_allocated() with is_swiotlb_active(), per Catalin]
Link: https://lkml.kernel.org/r/20251010173009.3916215-1-m.szyprowski@samsung.com
Link: https://lkml.kernel.org/r/20251009141508.2342138-1-m.szyprowski@samsung.com
Fixes: 370645f41e ("dma-mapping: force bouncing if the kmalloc() size is not cache-line-aligned")
Signed-off-by: Marek Szyprowski <m.szyprowski@samsung.com>
Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Inki Dae <m.szyprowski@samsung.com>
Cc: Robin Murohy <robin.murphy@arm.com>
Cc: "Isaac J. Manjarres" <isaacmanjarres@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Patch series "mm/damon/sysfs: fix commit test damon_ctx [de]allocation".
DAMON sysfs interface dynamically allocates and uses a damon_ctx object
for testing if given inputs for online DAMON parameters update is valid.
The object is being used without an allocation failure check, and leaked
when the test succeeds. Fix the two bugs.
This patch (of 2):
The damon_ctx for testing online DAMON parameters commit inputs is used
without its allocation failure check. This could result in an invalid
memory access. Fix it by directly returning an error when the allocation
failed.
Link: https://lkml.kernel.org/r/20251003201455.41448-1-sj@kernel.org
Link: https://lkml.kernel.org/r/20251003201455.41448-2-sj@kernel.org
Fixes: 4c9ea539ad ("mm/damon/sysfs: validate user inputs from damon_sysfs_commit_input()")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> [6.15+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Both the read and write side use kiocb_done() to finish a request, and
kiocb_done() will call io_put_kbuf() in case a provided buffer was used
for the request. Provided buffers are not supported for writes, hence
NULL is being passed in. This normally works fine, as io_put_kbuf()
won't actually use the value unless REQ_F_BUFFER_RING or
REQ_F_BUFFER_SELECTED is set in the request flags. But depending on
compiler (or whether or not CONFIG_CC_OPTIMIZE_FOR_SIZE is set), that
may be done even though the value is never used. This will then cause a
NULL pointer dereference.
Make it a bit more obvious and check for a NULL io_br_sel, and don't
even bother calling io_put_kbuf() for that case.
Fixes: 5fda512554 ("io_uring/kbuf: switch to storing struct io_buffer_list locally")
Reported-by: David Howells <dhowells@redhat.com>
Tested-by: David Howells <dhowells@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The S5_RESET_STATUS register is parsed on boot and printed to kmsg.
However, this could sometimes be misleading and lead to users wasting a
lot of time on meaningless debugging for two reasons:
* Some bits are never cleared by hardware. It's the software's
responsibility to clear them as per the Processor Programming Reference
(see [1]).
* Some rare hardware-initiated platform resets do not update the
register at all.
In both cases, a previous reboot could leave its trace in the register,
resulting in users seeing unrelated reboot reasons while debugging random
reboots afterward.
Write the read value back to the register in order to clear all reason bits
since they are write-1-to-clear while the others must be preserved.
[1]: https://bugzilla.kernel.org/show_bug.cgi?id=206537#attach_303991
[ bp: Massage commit message. ]
Fixes: ab81310287 ("x86/CPU/AMD: Print the reason for the last reset")
Signed-off-by: Rong Zhang <i@rong.moe>
Signed-off-by: Borislav Petkov (AMD) <bp@alien8.de>
Reviewed-by: Mario Limonciello (AMD) <superm1@kernel.org>
Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>
Cc: <stable@kernel.org>
Link: https://lore.kernel.org/all/20250913144245.23237-1-i@rong.moe/
devm_kasprintf() may return NULL on memory allocation failure,
but the debug message prints cpus->dai_name before checking it.
Move the dev_dbg() call after the NULL check to prevent potential
NULL pointer dereference.
Fixes: cb8ea62e64 ("ASoC: amd/sdw_utils: add sof based soundwire generic machine driver")
Signed-off-by: Li Qiang <liqiang01@kylinos.cn>
Link: https://patch.msgid.link/20251015075530.146851-1-liqiang01@kylinos.cn
Signed-off-by: Mark Brown <broonie@kernel.org>
Merge series from Haibo Chen <haibo.chen@nxp.com>:
PATCH 1: different operations maybe require different max frequency, so
add flexspi to handle such case, re-config the clock rate when
new coming operation require new clock frequency.
Patch 2: add workaround for erratum ERR050272. Since only add 4us dealy
in nxp_fspi_dll_calibration(), so do not distinguish different
platforms.
Patch 3: add max frequency limitation for different sample clock source
selection. Datasheet give max 66MHz for mode 0 and 166MHz for
mode 3. And IC suggest to add this limitation on all SoCs for
safety and stability.
We should wait for the rdma_cm to become SMBDIRECT_SOCKET_DISCONNECTED,
it turns out that (at least running some xfstests e.g. cifs/001)
often triggers the case where wait_event_interruptible() returns
with -ERESTARTSYS instead of waiting for SMBDIRECT_SOCKET_DISCONNECTED
to be reached.
Or we are already in SMBDIRECT_SOCKET_DISCONNECTING and never wait
for SMBDIRECT_SOCKET_DISCONNECTED.
Fixes: 050b8c3740 ("smbd: Make upper layer decide when to destroy the transport")
Fixes: e8b3bfe9bc ("cifs: smbd: Don't destroy transport on RDMA disconnect")
Fixes: b0aa92a229 ("smb: client: make sure smbd_disconnect_rdma_work() doesn't run after smbd_destroy() took over")
Cc: Steve French <smfrench@gmail.com>
Cc: Tom Talpey <tom@talpey.com>
Cc: Long Li <longli@microsoft.com>
Cc: Namjae Jeon <linkinjeon@kernel.org>
Cc: linux-cifs@vger.kernel.org
Cc: samba-technical@lists.samba.org
Signed-off-by: Stefan Metzmacher <metze@samba.org>
Signed-off-by: Steve French <stfrench@microsoft.com>
Eric Dumazet says:
====================
net: deal with sticky tx queues
Back in 2010, Tom Herbert added skb->ooo_okay to TCP flows.
Extend the feature to connected flows for other protocols like UDP.
skb->ooo_okay might never be set for bulk flows that always
have at least one skb in a qdisc queue of NIC queue,
especially if TX completion is delayed because of a stressed cpu
or aggressive interrupt mitigation.
The so-called "strange attractors" has caused many performance
issues, we need to do better now that TCP reacts better to
potential reorders.
Add new net.core.txq_reselection_ms sysctl to let
flows follow XPS and select a more efficient queue.
After this series, we no longer have to make sure threads
are pinned to cpus, they can migrate without adding
too much [spinlock, qdisc, TX completion] pressure anymore.
====================
Link: https://patch.msgid.link/20251013152234.842065-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
This is a followup of commit 726e9e8b94 ("tcp: refine
skb->ooo_okay setting") and of prior commit in this series
("net: control skb->ooo_okay from skb_set_owner_w()")
skb->ooo_okay might never be set for bulk flows that always
have at least one skb in a qdisc queue of NIC queue,
especially if TX completion is delayed because of a stressed cpu.
The so-called "strange attractors" has caused many performance
issues (see for instance 9b462d02d6 ("tcp: TCP Small Queues
and strange attractors")), we need to do better.
We have tried very hard to avoid reorders because TCP was
not dealing with them nicely a decade ago.
Use the new net.core.txq_reselection_ms sysctl to let
flows follow XPS and select a more efficient queue.
After this patch, we no longer have to make sure threads
are pinned to cpus, they now can be migrated without
adding too much spinlock/qdisc/TX completion pressure anymore.
TX completion part was problematic, because it added false sharing
on various socket fields, but also added false sharing and spinlock
contention in mm layers. Calling skb_orphan() from ndo_start_xmit()
is not an option unfortunately.
Note for later:
1) move sk->sk_tx_queue_mapping closer
to sk_tx_queue_mapping_jiffies for better cache locality.
2) Study if 9b462d02d6 ("tcp: TCP Small Queues
and strange attractors") could be revised.
Tested:
Used a host with 32 TX queues, shared by groups of 8 cores.
XPS setup :
echo ff >/sys/class/net/eth1/queue/tx-0/xps_cpus
echo ff00 >/sys/class/net/eth1/queue/tx-1/xps_cpus
echo ff0000 >/sys/class/net/eth1/queue/tx-2/xps_cpus
echo ff000000 >/sys/class/net/eth1/queue/tx-3/xps_cpus
echo ff,00000000 >/sys/class/net/eth1/queue/tx-4/xps_cpus
echo ff00,00000000 >/sys/class/net/eth1/queue/tx-5/xps_cpus
echo ff0000,00000000 >/sys/class/net/eth1/queue/tx-6/xps_cpus
echo ff000000,00000000 >/sys/class/net/eth1/queue/tx-7/xps_cpus
...
Launched a tcp_stream with 15 threads and 1000 flows, initially affined to core 0-15
taskset -c 0-15 tcp_stream -T15 -F1000 -l1000 -c -H target_host
Checked that only queues 0 and 1 are used as instructed by XPS :
tc -s qdisc show dev eth1|grep backlog|grep -v "backlog 0b 0p"
backlog 123489410b 1890p
backlog 69809026b 1064p
backlog 52401054b 805p
Then force each thread to run on cpu 1,9,17,25,33,41,49,57,65,73,81,89,97,105,113,121
C=1;PID=`pidof tcp_stream`;for P in `ls /proc/$PID/task`; do taskset -pc $C $P; C=$(($C + 8));done
Set txq_reselection_ms to 1000
echo 1000 > /proc/sys/net/core/txq_reselection_ms
Check that the flows have migrated nicely:
tc -s qdisc show dev eth1|grep backlog|grep -v "backlog 0b 0p"
backlog 130508314b 1916p
backlog 8584380b 126p
backlog 8584380b 126p
backlog 8379990b 123p
backlog 8584380b 126p
backlog 8487484b 125p
backlog 8584380b 126p
backlog 8448120b 124p
backlog 8584380b 126p
backlog 8720640b 128p
backlog 8856900b 130p
backlog 8584380b 126p
backlog 8652510b 127p
backlog 8448120b 124p
backlog 8516250b 125p
backlog 7834950b 115p
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20251013152234.842065-5-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
15 years after Tom Herbert added skb->ooo_okay, only TCP transport
benefits from it.
We can support other transports directly from skb_set_owner_w().
If no other TX packet for this socket is in a host queue (qdisc, NIC queue)
there is no risk of self-inflicted reordering, we can set skb->ooo_okay.
This allows netdev_pick_tx() to choose a TX queue based on XPS settings,
instead of reusing the queue chosen at the time the first packet was sent
for connected sockets.
Tested:
500 concurrent UDP_RR connected UDP flows, host with 32 TX queues,
512 cpus, XPS setup.
super_netperf 500 -t UDP_RR -H <host> -l 1000 -- -r 100,100 -Nn &
This patch saves between 10% and 20% of cycles, depending on how
process scheduler migrates threads among cpus.
Using following bpftrace script, we can see the effect on Qdisc/NIC tx queues
being better used (less cache line misses).
bpftrace -e '
k:__dev_queue_xmit { @start[cpu] = nsecs; }
kr:__dev_queue_xmit {
if (@start[cpu]) {
$delay = nsecs - @start[cpu];
delete(@start[cpu]);
@__dev_queue_xmit_ns = hist($delay);
}
}
END { clear(@start); }'
Before:
@__dev_queue_xmit_ns:
[128, 256) 6 | |
[256, 512) 116283 | |
[512, 1K) 1888205 |@@@@@@@@@@@ |
[1K, 2K) 8106167 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@ |
[2K, 4K) 8699293 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[4K, 8K) 2600676 |@@@@@@@@@@@@@@@ |
[8K, 16K) 721688 |@@@@ |
[16K, 32K) 122995 | |
[32K, 64K) 10639 | |
[64K, 128K) 119 | |
[128K, 256K) 1 | |
After:
@__dev_queue_xmit_ns:
[128, 256) 3 | |
[256, 512) 651112 |@@ |
[512, 1K) 8109938 |@@@@@@@@@@@@@@@@@@@@@@@@@@ |
[1K, 2K) 16081031 |@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@@|
[2K, 4K) 2411692 |@@@@@@@ |
[4K, 8K) 98994 | |
[8K, 16K) 1536 | |
[16K, 32K) 587 | |
[32K, 64K) 2 | |
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@google.com>
Link: https://patch.msgid.link/20251013152234.842065-3-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Some applications uses TCP_TX_DELAY socket option after TCP flow
is established.
Some metrics need to be updated, otherwise TCP might take time to
adapt to the new (emulated) RTT.
This patch adjusts tp->srtt_us, tp->rtt_min, icsk_rto
and sk->sk_pacing_rate.
This is best effort, and for instance icsk_rto is reset
without taking backoff into account.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Link: https://patch.msgid.link/20251013145926.833198-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Breno Leitao
Jakub Kicinski
Sabrina Dubroca
Eric Dumazet
Alexey Simakov
Justin Iurman
Pradeep Kumar Chitrapu
Baochen Qiang
Thiraviyam Mariyappan
Aditya Kumar Singh
Sarika Sharma
Jeff Johnson
Kang Yang
Dr. David Alan Gilbert
Linus Torvalds
Qiuxu Zhuo
Deepanshu Kartikey
Alexei Starovoitov
Marek Szyprowski
SeongJae Park
Lance Yang
Jens Axboe
Rong Zhang
Li Qiang
Sharique Mohammad
Frank Li
Mark Brown
Ankan Biswas
Stefan Metzmacher
Heiner Kallweit
Abhishek Rawal
Tim Hostetler