| Age | Commit message (Collapse) | Author |
|---|
|
iommufd_hwpt_replace_device() calls:
iommufd_auto_response_faults(hwpt, old_handle);
passing the *new* hwpt together with the handle of
the device's *old* domain. This should be a parameter mismatch:
1. Semantically, iommufd_auto_response_faults(x, handle) scans
x->fault's deliver list and response xarray for groups matching
"handle". A group is queued under the hwpt that was attached at
fault-delivery time. old_handle is fetched *before* the domain switch,
so its group lives on old->fault, not on the new hwpt->fault.
2. Historically, the first argument was "old". The routine was
introduced by commit b7d8833677ba ("iommufd: Fault-capable hwpt
attach/detach/replace") as __fault_domain_replace_dev() in
fault.c, correctly calling iommufd_auto_response_faults(old, curr).
Commit fb21b1568ada ("iommufd: Make attach_handle generic than
fault specific") moved this into iommufd_hwpt_replace_device() in
device.c and swapped it to "hwpt". This should be a refactor regression,
not an intentional change.
Fix this by passing "old" instead.
Link: https://patch.msgid.link/r/9D652384339C69D5+20260710122952.885325-1-peiyang_he@smail.nju.edu.cn
Fixes: fb21b1568ada ("iommufd: Make attach_handle generic than fault specific")
Cc: stable@vger.kernel.org
Signed-off-by: Peiyang He <peiyang_he@smail.nju.edu.cn>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
worker
A probe-time deadlock can occur between the dequeue worker and
drm_client_register(). During probe, drm_client_register() holds
clientlist_mutex and calls the fbdev hotplug callback, which triggers an
atomic commit that ends up sleeping in virtio_gpu_queue_ctrl_sgs()
waiting for virtqueue space. The dequeue worker that would free that
space calls virtio_gpu_cmd_get_display_info_cb(), which invokes
drm_kms_helper_hotplug_event() -> drm_client_dev_hotplug(), attempting
to acquire the same clientlist_mutex. Since wake_up() is only called
after the resp_cb loop, the probe thread is never woken and both threads
deadlock.
Fix this by removing the hotplug notification from
virtio_gpu_cmd_get_display_info_cb(). The display data (outputs[i].info)
is still updated synchronously in the callback.
For the init path, drm_client_register() already fires an initial
hotplug when the client is registered, which picks up the connector
state updated by display_info_cb.
For the runtime config_changed path, add a wait_event_timeout() in
config_changed_work_func() so that display_info_cb updates the connector
data before the hotplug notification is sent. Also replace
drm_helper_hpd_irq_event() with drm_kms_helper_hotplug_event() since
virtio-gpu never calls drm_kms_helper_poll_init() and thus
drm_helper_hpd_irq_event() always returns false without doing anything.
Fixes: 27655b9bb9f0 ("drm/client: Send hotplug event after registering a client")
Closes: https://syzkaller.appspot.com/bug?id=d6dd6f86d3aaf7eebe7406e45c1c6e549453f224
Closes: https://syzkaller.appspot.com/bug?id=908bd910da5dd79b88de4cf7baf376cc873a922e
Suggested-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Signed-off-by: Ryosuke Yasuoka <ryasuoka@redhat.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Link: https://patch.msgid.link/20260713-virtiogpu_syzbot-v2-1-2958fa37d46d@redhat.com
|
|
If start_streaming fails, then all queued buffers must be
returned to vb2 in state QUEUED.
Otherwise it will trigger a WARN_ON.
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Tested-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
In em28xx_free_device() dev_info passed &dev->intf->dev,
but that device can be freed already.
Just use pr_info instead.
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Tested-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
Use vb2_video_unregister_device instead of video_unregister_device
to ensure any streaming is correctly stopped at unregister time.
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Tested-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
Drop the em28xx_v4l2 'users' field, use v4l2_fh_is_singular_file()
instead.
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Tested-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
The em28xx driver creates a lot of video devices, but life-time management
is really bad. Instead use the struct v4l2_device release() callback to
have a single place where memory can be freed once the last user has gone.
Assisted-by: Claude:claude-opus-4-7
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Tested-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
When my em28xx USB device that uses the i2c tvp5150 driver is
disconnected, it crashes.
The cause is that the tvp5150 i2c module uses v4l2_async, but
the em28xx driver does not since it predates v4l2_async.
In that corner case sd->asc_list is empty, so
v4l2_async_unregister_subdev() never calls v4l2_device_unregister_subdev().
Modify the code so that, if sd->asc_list is empty,
v4l2_device_unregister_subdev() is still called.
Fixes: 28a1295795d8 ("media: v4l: async: Allow multiple connections between entities")
Signed-off-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Acked-by: Sakari Ailus <sakari.ailus@linux.intel.com>
Tested-by: Hans Verkuil <hverkuil+cisco@kernel.org>
Signed-off-by: Mauro Carvalho Chehab <mchehab+huawei@kernel.org>
|
|
Applies the same treatment as commit 7cf6dd467e87 ("drm/virtio:
Don't attach GEM to a non-created context in gem_object_open()")
to virtio_gpu_gem_object_close() to avoid trying to detach
a resource that was never attached due to a context
never being created when context_init is supported.
Fixes: 086b9f27f0ab ("drm/virtio: Don't create a context with default param if context_init is supported")
Cc: <stable@vger.kernel.org> # v6.14+
Signed-off-by: Jason Macnak <natsu@google.com>
Signed-off-by: Dmitry Osipenko <dmitry.osipenko@collabora.com>
Link: https://patch.msgid.link/20260625170828.3335431-1-natsu@google.com
|
|
Add error handling to devm_kasprintf functions in imx9_soc_probe().
Assisted-by: gkh_clanker_2000
Cc: Frank Li <Frank.Li@nxp.com>
Cc: Sascha Hauer <s.hauer@pengutronix.de>
Cc: Pengutronix Kernel Team <kernel@pengutronix.de>
Cc: Fabio Estevam <festevam@gmail.com>
Cc: Peng Fan <peng.fan@nxp.com>
Cc: Shawn Guo <shawnguo@kernel.org>
Cc: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Signed-off-by: Griffin Kroah-Hartman <griffin@kroah.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Frank Li <Frank.Li@nxp.com>
|
|
When dma_buf_export() succeeds but dma_buf_fd() fails (e.g. -EMFILE from
fd exhaustion), the dmabuf is leaked with no dma_buf_put() called.
Reproducer: exhaust fd table near RLIMIT_NOFILE, then repeatedly call
IOMMU_TEST_OP_DMABUF_GET — htop shows unbounded memory growth.
Fix by calling dma_buf_put(dmabuf) on error and returning directly.
Fixes: d2041f1f11dd ("iommufd/selftest: Add some tests for the dmabuf flow")
Link: https://patch.msgid.link/r/20260707030635.221577-1-seven.yi.lee@gmail.com
Signed-off-by: yeeli <seven.yi.lee@gmail.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Pull rdma fixes from Jason Gunthorpe:
"More bug fixes, many found by tools:
- Protect from a possible DOS with certain RMPP traffic patterns
- Correct mac address comparison so CMA works properly on IB
- Some crashes in irdma around memory registration
- Uninitialized value in erdma and mana
- Wrong order setting up a QP in SIW allowed a network packet to
reach an unready QP struct
- Catch math overflows in hns and irdma"
* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
RDMA/irdma: Prevent overflows in memory contiguity checks
RDMA/siw: publish QP after initialization
RDMA/hns: Fix potential integer overflow in mhop hem cleanup
RDMA/core: Fix memory leak in __ib_create_cq() on invalid cqe
RDMA/mana_ib: initialize err for empty send WR lists
RDMA/erdma: initialize ret for empty receive WR lists
RDMA/irdma: Prevent user-triggered null deref on QP create
RDMA/irdma: Prevent rereg_mr for non-mem regions
RDMA/cma: Fix hardware address comparison length in netevent callback
RDMa/mlx5: Avoid frame overflow warning
IB/mad: Drop unmatched RMPP responses before reassembly
|
|
On most Qualcomm SoCs where minidump is supported, a word in always-on
SRAM is shared between the operating system (OS) and boot firmware.
Before DDR is initialized on the warm reset following a crash, firmware
reads this word to decide if minidump is enabled and collect a minidump,
and where to deliver it (USB upload to a host, or save to local
storage). The OS is expected to select one of the destinations.
The SRAM region is described by a 'sram' phandle on the SCM DT node.
If the property is absent the feature is silently disabled, keeping
existing SoCs unaffected.
Expose a 'minidump_dest' module parameter (default: usb) so the user
can select the destination. Only the string names "usb" or "storage"
are accepted.
Reviewed-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Signed-off-by: Mukesh Ojha <mukesh.ojha@oss.qualcomm.com>
Link: https://lore.kernel.org/r/20260624190830.3131112-6-mukesh.ojha@oss.qualcomm.com
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
|
|
Replace the bare `return ret` after qcom_scm_find_dload_address() with
dev_err_probe() to produce a consistent, deferred-probe-aware error
message when the download-mode address cannot be resolved.
Reviewed-by: Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Signed-off-by: Mukesh Ojha <mukesh.ojha@oss.qualcomm.com>
Link: https://lore.kernel.org/r/20260624190830.3131112-5-mukesh.ojha@oss.qualcomm.com
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
|
|
__scm is published in qcom_scm_probe() with smp_store_release()
but qcom_scm_set_download_mode() reads it directly without
smp_load_acquire(), creating a potential ordering violation where a
CPU could observe a stale or partially initialised __scm pointer.
Use smp_load_acquire() to acquire __scm in the
qcom_scm_set_download_mode() function.
Signed-off-by: Mukesh Ojha <mukesh.ojha@oss.qualcomm.com>
Link: https://lore.kernel.org/r/20260624190830.3131112-4-mukesh.ojha@oss.qualcomm.com
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
|
|
Switch drm/msm client drivers over to generic PAS TZ APIs. Generic PAS
TZ service allows to support multiple TZ implementation backends like QTEE
based SCM PAS service, OP-TEE based PAS service and any further future TZ
backend service.
Acked-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>
Reviewed-by: Mukesh Ojha <mukesh.ojha@oss.qualcomm.com>
Tested-by: Mukesh Ojha <mukesh.ojha@oss.qualcomm.com> # Lemans
Signed-off-by: Sumit Garg <sumit.garg@oss.qualcomm.com>
Link: https://lore.kernel.org/r/20260702115835.167602-9-sumit.garg@kernel.org
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
|
|
qca_controller_memdump() allocates qca->qca_memdump before processing
the first dump packet. For a sequence-zero packet it then disables IBS,
marks memdump collection active, and reads the advertised dump size.
If the controller reports a zero dump size, the error path frees the
local qca_memdump object and returns without clearing qca->qca_memdump
or undoing the collection state. A later memdump work item initializes
its local pointer from qca->qca_memdump and skips allocation when that
pointer is non-NULL, so it can operate on freed memory. The stale
collection and IBS-disabled flags can also leave waiters or later
transmit handling blocked behind an aborted dump.
Clear the saved pointer and memdump state before returning from the
invalid-size path, matching the cleanup used when hci_devcd_init() fails.
A static analysis checker reported the stale memdump state, and manual
source review confirmed the invalid-size failure path.
Fixes: 06d3fdfcdf5c ("Bluetooth: hci_qca: Add qcom devcoredump support")
Signed-off-by: Ruoyu Wang <ruoyuw560@gmail.com>
Reviewed-by: Paul Menzel <pmenzel@molgen.mpg.de>
Reviewed-by: Zijun Hu <zijun.hu@oss.qualcomm.com>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
|
|
rtlbt_parse_firmware() copies patch_length - 4 bytes before appending the
firmware version. A malformed firmware patch shorter than the version field
can make this subtraction underflow and turn the copy into an oversized
read and write during Bluetooth setup.
The existing patch_offset + patch_length check can also wrap on 32-bit
architectures. Validate the patch length and range without arithmetic
overflow before allocating or copying the patch.
Fixes: db33c77dddc2 ("Bluetooth: btrtl: Create separate module for Realtek BT driver")
Cc: stable@vger.kernel.org
Signed-off-by: Laxman Acharya Padhya <acharyalaxman8848@gmail.com>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
|
|
In the TLV_TYPE_NVM branch of qca_tlv_check_data() the tag loop bound is
"while (idx < length - sizeof(struct tlv_type_nvm))". "length" is a signed
int from the firmware TLV header and sizeof(struct tlv_type_nvm) is a
size_t (12), so "length" is converted to size_t and any firmware-supplied
"length" < 12 makes the subtraction wrap to a huge value. The loop body
then reads a 12-byte struct tlv_type_nvm past the end of the short
vmalloc'd firmware buffer (and the EDL_TAG_ID_* handlers can write past it).
Rewrite the bound as "idx + sizeof(struct tlv_type_nvm) <= length"; both
operands are non-negative, so it no longer underflows and a "length" too
small for one record correctly skips the loop.
BUG: KASAN: vmalloc-out-of-bounds in qca_download_firmware.isra.0 (drivers/bluetooth/btqca.c:421)
Read of size 2 at addr ffffc900000e5004 by task kworker/u9:0/52
Workqueue: hci0 hci_power_on
Call Trace:
...
kasan_report (mm/kasan/report.c:595)
qca_download_firmware.isra.0 (drivers/bluetooth/btqca.c:421 drivers/bluetooth/btqca.c:617)
qca_uart_setup (drivers/bluetooth/btqca.c:948)
qca_setup (drivers/bluetooth/hci_qca.c:2029)
hci_uart_setup (drivers/bluetooth/hci_ldisc.c:438)
hci_dev_open_sync (net/bluetooth/hci_sync.c:5227)
hci_power_on (net/bluetooth/hci_core.c:920)
process_one_work (kernel/workqueue.c:3322)
worker_thread (kernel/workqueue.c:3486)
kthread (kernel/kthread.c:436)
ret_from_fork (arch/x86/kernel/process.c:158)
ret_from_fork_asm (arch/x86/entry/entry_64.S:245)
Fixes: 2e4edfa1e2bd ("Bluetooth: qca: add missing firmware sanity checks")
Reported-by: Weiming Shi <bestswngs@gmail.com>
Assisted-by: Claude:claude-opus-4-8
Signed-off-by: Xiang Mei <xmei5@asu.edu>
Reported-by: Weiming Shi <bestswngs@gmail.com>
Reviewed-by: Johan Hovold <johan@kernel.org>
Acked-by: Bartosz Golaszewski <bartosz.golaszewski@oss.qualcomm.com>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
|
|
The failover continuation in carl9170_rx_stream() copies the full tlen
from the second USB transfer instead of capping at rx_failover_missing
bytes. When both transfers are near maximum size, the total exceeds the
65535-byte failover SKB, triggering skb_over_panic.
Limit the copy size to the missing byte count.
Fixes: a84fab3cbfdc ("carl9170: 802.11 rx/tx processing and usb backend")
Signed-off-by: Tristan Madani <tristan@talencesecurity.com>
Acked-by: Christian Lamparter <chunkeey@gmail.com>
Closes: https://syzkaller.appspot.com/bug?extid=5c1ca6ccaa1215781cac
Link: https://patch.msgid.link/20260421134929.325662-4-tristmd@gmail.com
[Fix checkpatch CHECK:PARENTHESIS_ALIGNMENT]
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
The bounds check in carl9170_tx_process_status() uses
`i > ((cmd->hdr.len / 2) + 1)` which is off by two, allowing
2 extra iterations past valid _tx_status entries when the firmware-
controlled hdr.ext exceeds hdr.len/2. Fix by using the correct
comparison `i >= (cmd->hdr.len / 2)`.
Fixes: a84fab3cbfdc ("carl9170: 802.11 rx/tx processing and usb backend")
Signed-off-by: Tristan Madani <tristan@talencesecurity.com>
Acked-by: Christian Lamparter <chunkeey@gmail.com>
Closes: https://syzkaller.appspot.com/bug?extid=5c1ca6ccaa1215781cac
Link: https://patch.msgid.link/20260421134929.325662-3-tristmd@gmail.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
When the firmware sends a command response with a length mismatch,
carl9170_cmd_callback() logs the mismatch and calls carl9170_restart()
but then falls through to memcpy(ar->readbuf, buffer + 4, len - 4).
Since len comes from the firmware and can exceed ar->readlen, this
copies more data than the readbuf was allocated for.
Bound the memcpy to min(len - 4, ar->readlen) so that the response
is still completed -- avoiding repeated restarts from queued garbage --
while preventing an overread past the response buffer.
Fixes: a84fab3cbfdc ("carl9170: 802.11 rx/tx processing and usb backend")
Signed-off-by: Tristan Madani <tristan@talencesecurity.com>
Acked-by: Christian Lamparter <chunkeey@gmail.com>
Closes: https://syzkaller.appspot.com/bug?extid=5c1ca6ccaa1215781cac
Link: https://patch.msgid.link/20260421134929.325662-2-tristmd@gmail.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
The firmware-controlled beacon_ie_len, assoc_req_len, and assoc_resp_len
fields in ath6kl_wmi_connect_event_rx() are not validated against the
buffer length. Their sum (up to 765) can exceed the actual WMI event
data, causing out-of-bounds reads during IE parsing and state corruption
of wmi->is_wmm_enabled.
Add a check that the total IE length fits within the buffer.
Fixes: bdcd81707973 ("Add ath6kl cleaned up driver")
Signed-off-by: Tristan Madani <tristan@talencesecurity.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20260421135009.348084-3-tristmd@gmail.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
The firmware-controlled num_msg field (u8, 0-255) drives the loop in
ath6kl_wmi_tx_complete_event_rx() without validation against the buffer
length. This allows out-of-bounds reads of up to 1020 bytes past the
WMI event buffer when the firmware sends an inflated num_msg.
Add a check that the buffer is large enough to hold the fixed struct
and the num_msg variable-length entries.
Fixes: bdcd81707973 ("Add ath6kl cleaned up driver")
Signed-off-by: Tristan Madani <tristan@talencesecurity.com>
Link: https://patch.msgid.link/20260625232907.3620746-1-tristmd@gmail.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
aggr_recv_addba_req_evt() logs a debug message when the firmware-supplied
win_sz is outside [AGGR_WIN_SZ_MIN, AGGR_WIN_SZ_MAX] but does not
return. The out-of-range win_sz is then used in TID_WINDOW_SZ() to
compute a kzalloc size and stored in rxtid->hold_q_sz, leading to
zero-size or overflowed allocations and subsequent out-of-bounds access.
Clean up any previously active aggregation session for the TID first,
then return early when win_sz is out of the valid range, instead of
proceeding with a broken allocation size.
Fixes: bdcd81707973 ("Add ath6kl cleaned up driver")
Cc: stable@vger.kernel.org
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Signed-off-by: Tristan Madani <tristan@talencesecurity.com>
Link: https://patch.msgid.link/20260702005020.708717-1-tristmd@gmail.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
When unbinding the ath12k driver, kernel NULL pointer dereferences
occur in irq_work_sync() called from rhashtable_destroy().
Two hash tables are affected:
1. ath12k_link_sta hash table in ath12k_base
2. ath12k_dp_link_peer hash table in ath12k_dp
The issue happens because the destroy functions are called unconditionally
in cleanup paths, but the hash tables are only initialized late in their
respective init functions. If the device was never fully started or if the
init functions failed before initializing the hash tables, the pointers
will be NULL. The issues are always reproducible from a VM because the MSI
addressing initialization is failing.
Call trace for ath12k_link_sta_rhash_tbl_destroy:
RIP: irq_work_sync+0x1e/0x70
rhashtable_destroy+0x12/0x60
ath12k_link_sta_rhash_tbl_destroy+0x19/0x40 [ath12k]
ath12k_core_stop+0xe/0x80 [ath12k]
ath12k_core_hw_group_cleanup+0x6b/0xb0 [ath12k]
ath12k_pci_remove+0x60/0x110 [ath12k]
Call trace for ath12k_dp_link_peer_rhash_tbl_destroy:
RIP: irq_work_sync+0x1e/0x70
rhashtable_destroy+0x12/0x60
ath12k_dp_link_peer_rhash_tbl_destroy+0x29/0x50 [ath12k]
ath12k_dp_cmn_device_deinit+0x21/0x140 [ath12k]
ath12k_core_hw_group_cleanup+0x6b/0xb0 [ath12k]
ath12k_pci_remove+0x60/0x110 [ath12k]
Fix this by adding NULL checks before calling rhashtable_destroy() in
both destroy functions.
The NULL check approach was chosen because the rhashtable pointer
serves as the initialization state indicator. The init can fail at
various points, leaving some components uninitialized. Checking the
pointer directly is simpler than adding separate state flags that
would need synchronization.
Fixes: 57ccca410237 ("wifi: ath12k: Add hash table for ath12k_link_sta in ath12k_base")
Fixes: a88cf5f71adf ("wifi: ath12k: Add hash table for ath12k_dp_link_peer")
Cc: stable@vger.kernel.org
Signed-off-by: Jose Ignacio Tornos Martinez <jtornosm@redhat.com>
Reviewed-by: Vasanthakumar Thiagarajan <vasanthakumar.thiagarajan@oss.qualcomm.com>
Link: https://patch.msgid.link/20260615112103.601982-1-jtornosm@redhat.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
Commit [1] introduced a regression causing severely degraded MLO RX
throughput on WCN7850.
On WCN7850, there is only a single ar instance, but MLO uses two
link IDs. ath12k_dp_peer->hw_links[] is indexed using ar->hw_link_id,
which causes both MLO link IDs to be stored at the same index.
As a result, an incorrect link ID is assigned to MSDUs in
ath12k_dp_rx_deliver_msdu(), leading to severe MLO RX throughput loss.
Different chipsets identify the per-MSDU link differently:
- On QCN9274 / IPQ5332, the host owns multiple ar instances and the
per-MSDU hw_link_id from the RX descriptor maps cleanly through
dp_peer->hw_links[hw_link_id] to the IEEE link_id.
- On single-ar chipsets like WCN7850 / QCC2072, there is only one ar
instance for both MLO links, so dp_peer->hw_links[] has just one
valid slot and cannot be used to distinguish the two links. To
resolve the link, walk dp_peer->link_peers[] and match by
rxcb->peer_id, which on the link_peer side identifies the link
peer for the MSDU.
Add a new hw_op set_rx_link_id() so each chipset resolves the link
on the RX fast path using whatever signal it actually has, and let
the op itself decide whether to populate rx_status::link_valid and
rx_status::link_id:
QCN9274 / IPQ5332 : always derive link_id from
dp_peer->hw_links[rxcb->hw_link_id] and set
link_valid.
WCN7850 / QCC2072 : walk the link_peers[] of dp_peer to find the
link_peer whose peer_id matches rxcb->peer_id,
and set link_valid only when a match is found.
Otherwise leave link_valid clear so that
mac80211 can fall back to its own link
resolution path (via addr2 / deflink).
For WCN7850 / QCC2072, walking dp_peer->link_peers[] is bounded by
the number of links actually populated, so introduce a link_peers_map
bitmap (unsigned long) in struct ath12k_dp_peer that tracks populated
slots and use for_each_set_bit() to iterate. Non-MLO clients hit one
slot, current MLO clients hit two; the full ATH12K_NUM_MAX_LINKS
array is never scanned. The bitmap is maintained with WRITE_ONCE() on
the write side (under dp_hw->peer_lock) paired with READ_ONCE() on
both the lockless RX read side and the write-side RMW for KCSAN
correctness.
Also guard the dp_peer dereference in ath12k_mac_peer_cleanup_all()
with a NULL check, since peer->dp_peer can be NULL for self-peers or
peers not yet fully assigned, the pre-existing rcu_assign_pointer()
call there had the same latent issue.
This restores the correct link ID on WCN7850 without changing the
QCN9274 / IPQ5332 data path, which keeps its O(1) hw_links[]
indexing.
Tested-on: WCN7850 hw2.0 PCI WLAN.HMT.1.1.c5-00302-QCAHMTSWPL_V1.0_V2.0_SILICONZ-1.115823.3
Fixes: 11157e0910fd ("wifi: ath12k: Use ath12k_dp_peer in per packet Tx & Rx paths") # [1]
Signed-off-by: Yingying Tang <yingying.tang@oss.qualcomm.com>
Reviewed-by: Baochen Qiang <baochen.qiang@oss.qualcomm.com>
Reviewed-by: Rameshkumar Sundaram <rameshkumar.sundaram@oss.qualcomm.com>
Link: https://patch.msgid.link/20260610053315.2249912-1-yingying.tang@oss.qualcomm.com
Signed-off-by: Jeff Johnson <jeff.johnson@oss.qualcomm.com>
|
|
IOMMU_TEST_OP_DIRTY sizes its temporary dirty bitmap from length /
page_size. Very large selftest ranges can make the DIV_ROUND_UP()
additions wrap before allocation, producing a zero-length allocation while
the later test_bit() loop still walks the original number of bits.
The selftest helper does not need to support unbounded dirty bitmap sizes.
Reject requests that would allocate more than SZ_16M for the temporary
buffer.
Fixes: 79ea4a496ab5 ("iommufd/selftest: Fix buffer read overrrun in the dirty test")
Link: https://patch.msgid.link/r/20260628152331.82122.408afd7b466c.iommufd-test-dirty-bitmap-size-wrap@trailofbits.com
Assisted-by: Codex:gpt-5.5-cyber-preview
Signed-off-by: Samuel Moelius <sam.moelius@trailofbits.com>
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
|
|
Renesas RZ/G3S SoC has a CAN-FD IP. Add clocks and resets for it.
Reviewed-by: Biju Das <biju.das.jz@bp.renesas.com>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Claudiu Beznea <claudiu.beznea.uj@bp.renesas.com>
Link: https://patch.msgid.link/20260709182332.876408-2-claudiu.beznea+renesas@tuxon.dev
Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
|
|
init_one() defers handling errors from hfi1_init() and
hfi1_register_ib_device() to a combined block. This allows IB registration
to run after device initialization has failed. Cleanup then depends on two
unrelated error values.
The late probe failure path also differs from the common teardown in
remove_one(), even though both release the same initialized hardware and
driver resources. Maintaining separate sequences obscures ownership and
allows the paths to drift whenever initialization changes.
Unwind at each failing stage and keep late probe teardown ordered like
remove_one(), so partial initialization releases exactly the resources it
owns and normal removal remains the reference cleanup flow.
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-13-b9e9641268a5@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
Commit ed6f653fe430 ("staging/rdma/hfi1: Fix debugfs access race") moved
debugfs creation after device initialization and IB registration so users
cannot access the files before the driver is ready. However, init_one()
still creates them before character device creation and SDMA startup
finish.
Move hfi1_dbg_ibdev_init() to the end of the successful probe path,
matching hfi1_dbg_ibdev_exit() as the first action in remove_one().
Fixes: ed6f653fe430 ("staging/rdma/hfi1: Fix debugfs access race")
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-12-b9e9641268a5@nvidia.com
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
init_one() creates the character device before checking whether generic or
IB initialization failed, only to remove it immediately while unwinding.
Moreover, user_add() already calls user_remove() when device creation
fails.
Move hfi1_device_create() after the initialization failure path,
immediately before starting SDMA. The failure path then has no character
device to remove, and hfi1_device_create() continues to unwind its own
failures.
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-11-b9e9641268a5@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
hfi1 does not queue work on ib_wq. QSFP and link work run on the per-port
link_wq, while the remaining device work uses hfi1_wq or dedicated queues.
The probe failure path destroys both per-port workqueues, and normal device
removal flushes them in shutdown_device() before destroying them.
Remove the flushes of the core-owned global workqueue. Waiting for
unrelated core or other device work is not part of hfi1 teardown.
Fixes: 71d47008ca1b ("IB/hfi1: Create workqueue for link events")
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-10-b9e9641268a5@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
create_workqueues() is called only from init_one(), immediately after
hfi1_alloc_devdata() returns a zero-initialized hfi1_devdata. As a
result, the per-port hfi1_wq and link_wq pointers are always NULL on
entry, and nothing modifies them between allocation and this call.
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-9-b9e9641268a5@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
After the preceding changes, module parameter validation and common PCI
setup no longer need hfi1_devdata. init_one() nevertheless allocates it
first, so failures in those early steps return without releasing it.
Move the allocation after hfi1_pcie_init() and return directly when no
resources are held. Use dev_err() and pci_info() for diagnostics emitted
before allocation so they retain the adapter BDF. Once PCI setup succeeds,
unwind allocation failures through hfi1_pcie_cleanup() to disable the
device and release its regions.
Some PCI error delivery paths are not serialized against probe. Keep their
diagnostics based on pci_dev and skip resume while driver data is absent,
preventing recovery from dereferencing a missing hfi1_devdata.
Fixes: 57f97e96625f ("IB/hfi1: Get the hfi1_devdata structure as early as possible")
Reported-by: Dawei Feng <dawei.feng@seu.edu.cn>
Closes: https://lore.kernel.org/all/20260627060159.2543686-1-dawei.feng@seu.edu.cn/
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-8-b9e9641268a5@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
hfi1_init_dd() allocates the shared AIP/VNIC RX support before returning.
If hfi1_init() or hfi1_register_ib_device() later fails, init_one() tears
down the device data without calling hfi1_free_rx(). This leaks netdev_rx
and its dummy netdev.
Free the RX support after IB unregistration and before postinit_cleanup(),
as done on normal device removal.
Fixes: 4730f4a6c6b2 ("IB/hfi1: Activate the dummy netdev")
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-7-b9e9641268a5@nvidia.com
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
create_workqueues() only needs fields set up by hfi1_alloc_devdata().
Call it before hfi1_init_dd() so a workqueue allocation failure happens
before chip resources are initialized.
To keep the reordered error paths safe, make init_one() own hfi1_devdata.
hfi1_init_dd() unwinds its partial setup but leaves the allocation for the
caller to free. If device initialization fails, destroy the workqueues
before freeing the device data.
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-6-b9e9641268a5@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
hfi1_validate_rcvhdrcnt() only needs hfi1_devdata to identify the adapter
in error messages. Requiring the full device data prevents module parameter
validation from running before hfi1_devdata is allocated.
Pass pci_dev instead and use dev_err(), allowing validation to move earlier
without losing the PCI BDF needed on multi-device systems. Use %u for the
unsigned count while changing the messages.
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-5-b9e9641268a5@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
hfi1_pcie_init() only needs hfi1_devdata to reach the PCI device. This
unnecessary dependency prevents common PCI setup from running before
hfi1_devdata is allocated.
Pass pci_dev directly and report failures with dev_err(), preserving the
device BDF needed to identify the failing adapter on multi-device systems.
Use %pe while changing the messages so errno values are decoded.
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-4-b9e9641268a5@nvidia.com
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
The PCI core calls init_one() only after pci_match_device() has selected
an ID. For normal probing, hfi1_pci_tbl already restricts matches to the
two supported Intel device IDs. Dynamic IDs and driver_override are
explicit requests to attempt binding, so the probe should not second-guess
the PCI core's decision.
Remove the redundant check.
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-3-b9e9641268a5@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
hfi1_free_devdata() assumes that the device was inserted into the unit
table and unconditionally erases dd->unit. If xa_alloc_irq() fails, the
zero-initialized unit remains zero, so full cleanup can remove an
unrelated device from index 0.
Release only the rdmavt allocation and return immediately while the unit
table has not acquired the device.
Fixes: 03b92789e5cf ("hfi1: Convert hfi1_unit_table to XArray")
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-2-b9e9641268a5@nvidia.com
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
rvt_alloc_device() deallocates the IB device when its port array cannot
be allocated but then returns the pointer to the released allocation.
Callers treat any non-NULL value as valid and dereference it, resulting
in a use-after-free.
Return NULL immediately after deallocation so callers can propagate the
allocation failure.
Fixes: ff6acd69518e ("IB/rdmavt: Add device structure allocation")
Link: https://patch.msgid.link/20260708-clean-init-one-hfi1-v1-1-b9e9641268a5@nvidia.com
Reviewed-by: Kalesh AP <kalesh-anakkur.purayil@broadcom.com>
Signed-off-by: Leon Romanovsky <leonro@nvidia.com>
|
|
fwnet_frag_new() keeps a sorted list of received fragments for a partial
datagram. When a new fragment is adjacent to an existing fragment, the
code checks whether the new fragment also closes the gap to the next or
previous list entry.
Those neighbor lookups currently assume that the current fragment always
has a real next or previous fragment. At a list edge, the next or
previous entry is the list head, not a struct fwnet_fragment_info.
The gap checks also compare against the old edge of the current fragment
instead of the edge after adding the new fragment. As a result, a
fragment that bridges two existing ranges may leave two adjacent ranges
unmerged, so fwnet_pd_is_complete() can miss a complete datagram.
Check for the list head before looking up the neighboring fragment, and
compare the neighbor against the new fragment's far edge when deciding
whether to merge all three ranges.
This issue was found by a static analysis checker and confirmed by
manual source review.
Fixes: c76acec6d551 ("firewire: add IPv4 support")
Signed-off-by: Ruoyu Wang <ruoyuw560@gmail.com>
Link: https://lore.kernel.org/r/20260707150454.2265951-1-ruoyuw560@gmail.com
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
|
|
Some Realtek USB SD readers with a tray keep raw SD_CD asserted
when an empty tray is inserted. The MMC core then repeatedly tries to
initialize non-existent media, sees command timeouts, calls ->get_cd()
again, and starts the same detect cycle over.
Do not qualify media by open-coding MMC commands in ->get_cd().
Instead, let the normal MMC rescan path probe the card. If an
initialization command times out before a card has been attached,
suppress the raw SD_CD signal so the host can settle and the USB
parent can autosuspend.
Clear the suppression only when raw SD_CD drops. On the affected tray
reader, changing media requires removing and reinserting the tray, so
a low CD transition is the signal that a new insertion attempt can be
trusted again.
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ulf Hansson <ulfh@kernel.org>
|
|
The runtime autosuspend callback currently queries card status and
clears OCP by issuing USB register accesses. This can run from the
USB runtime-PM path itself, which is the wrong place to start more
device I/O.
Keep a cached copy of the card-status bits from normal status reads
instead. During runtime autosuspend, use that cached value only to
preserve the existing Memory Stick autosuspend deferral.
Do not treat raw SD_CD as an autosuspend blocker, because tray-based
SD readers can assert SD_CD with an empty tray. A real SD card is
protected by the SD/MMC child runtime-PM usage once powered.
Also stop clearing OCP from the runtime autosuspend callback, so the
callback does not issue USB commands.
Fixes: bb400d2120bd ("mfd: rtsx_usb: Defer autosuspend while card exists")
Signed-off-by: Sean Rhodes <sean@starlabs.systems>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Ulf Hansson <ulfh@kernel.org>
|
|
Several SCMI notification data structures allocate a descriptor and then
allocate a fixed-size array referenced by that descriptor.
Use flexible array members for per-protocol registered events and
per-event source refcounts so each descriptor and its array can be
allocated together. Annotate the flexible arrays with __counted_by() and
assign the count fields immediately after allocation, before any array
access.
The registered_protocols table has a fixed SCMI_MAX_PROTO size, so embed
it directly in the notification instance instead of allocating it
separately.
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20260630204719.334211-1-rosenp@gmail.com
(sudeep.holla: Updated title and improved change description)
Signed-off-by: Sudeep Holla <sudeep.holla@kernel.org>
|
|
On multi-socket platforms each socket has its own BPMP that is
registered with the kernel. For such platforms prefix the NUMA ID for
each socket to the BPMP powergate name to ensure there is a unique name
for each power-domain. Note that we only add the NUMA ID for powergates
that return a valid name because an invalid name indicates that the
powergate ID is not supported.
Note the check for the NULL string is moved into the function
tegra_bpmp_powergate_get_name(), because in the multi-socket case we
must only add the prefix if we receive a valid name. A NULL string
indicates that there is no valid powergate associated with the ID that
is being queried.
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Ulf Hansson <ulfh@kernel.org>
|
|
Commit 18605b1b936b ("pmdomain: bcm: bcm2835-power: Increase ASB control
timeout") raised the ASB handshake polling budget from 1us to 5us.
Surveying the pmdomain subsystem, 5us is still one of the smallest polling
budgets by a wide margin. Comparable handshakes in other drivers use:
- 100us : starfive jh71xx-pmu, apple pmgr-pwrstate
- 1ms : renesas rcar-sysc, rmobile-sysc (power-on)
- 10ms : renesas rcar-gen4-sysc, sunxi sun55i-pck600
- 1s : mediatek mtk-pm-domains, mtk-scpsys
Raise the BCM2835 timeout to 100us, matching analogous drivers. 100us is
still negligible relative to a power-domain transition and gives the V3D
master ASB substantially more headroom to drain under heavy workloads,
assuring us that the timeout is enough for any scenario.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Acked-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: Ulf Hansson <ulfh@kernel.org>
|
|
Add support for the power domains provided by the HFRPSYS Power
Controller of the MT8196 SoC.
Those control power to the eDP and DP Transmitter IPs.
Reviewed-by: Matthias Brugger <matthias.bgg@gmail.com>
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Ulf Hansson <ulfh@kernel.org>
|
|
Some new SoCs like MT8196, MT6991, and others, have got one
additional power controller (usually in the HFRP Multimedia
block) which needs a simplified power on/off sequence while
using Direct Control strategy.
Domains using the "simple power sequence" are not backed by
the RTFF hardware, have no Bus Protection mechanism, lacks
the ISO, PWR_ON, PWR_ON_2ND bits, and therefore get enabled
automatically after getting out of reset.
This simple power sequence is then a subset of the full one
as only needs the enablement of the specific power domain's
clock input and reset (where, again, after getting out of
reset, the ISO and PWR_ON bits are automatically internally
getting flipped) to enable or disable (power on or off).
Moreover, the simple power sequence power domains guarantee
that they always get enabled/disabled after executing the
relevant power sequence (on/off) so, differently from the
others, there is also no need to poll for a PWR_ACK.
Reviewed-by: Matthias Brugger <matthias.bgg@gmail.com>
Signed-off-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Signed-off-by: Ulf Hansson <ulfh@kernel.org>
|
