- Don't try to power on/off devices unless we know they actually support
power control (Manivannan Sadhasivam)
* pci/pwrctrl:
PCI/pwrctrl: Lock device when calling device_is_bound()
PCI/pwrctrl: Do not try to power on/off devices that don't need pwrctrl
PCI/pwrctrl: Move pci_pwrctrl_is_required() earlier in file
- Fix race between pci_proc_init() and pci_bus_add_device() (Krzysztof
Wilczyński)
* pci/procfs:
PCI/proc: Fix race between pci_proc_init() and pci_bus_add_device()
- Set power state to 'unknown' for all devices, not just those with
drivers, during suspend (Lukas Wunner)
- Skip restoring Resizable BARs and VF Resizable BARs if device doesn't
respond to config reads, to avoid invalid array accesses (Marco
Nenciarini)
- Add pci_suspend_retains_context() so drivers can tell whether devices may
be reset while resuming from suspend due to platform issues; use this in
nvme to avoid issues on Qcom RCs (Manivannan Sadhasivam)
* pci/pm:
nvme-pci: Use pci_suspend_retains_context() during suspend
PCI: qcom: Indicate broken L1SS exit during resume from system suspend
PCI: Indicate context lost if L1SS exit is broken during resume from system suspend
PCI: Add pci_suspend_retains_context() to check if device state is preserved during suspend
PCI/IOV: Skip VF Resizable BAR restore on read error
PCI: Skip Resizable BAR restore on read error
PCI: Stop setting cached power state to 'unknown' on unbind
- Prevent P2PDMA as well as CPU access to non-mappable BARs, e.g., s390 ISM
BARs (Matt Evans)
- Add Intel QAT, DSA, IAA devices to whitelist (Lukas Wunner)
* pci/p2pdma:
PCI/P2PDMA: Add Intel QAT, DSA, IAA devices to whitelist
PCI/P2PDMA: Avoid returning a provider for non_mappable_bars
- Remove MPS/MRRS Kconfig settings (CONFIG_PCIE_BUS_*) that worked around a
WiFi device defect (Bjorn Helgaas)
- Always lift 2.5GT/s restriction in PCIe failed link retraining to avoid
clamping a link to 2.5GT/s after hot-plug changes the device (Maciej W.
Rozycki)
- Don't bother trying to retrain a 2.5GT/s link at 2.5GT/s since nothing
would be gained by the retrain (Maciej W. Rozycki)
* pci/enumeration:
PCI: Bail out early for 2.5GT/s devices in PCIe failed link retraining
PCI: Use pcie_get_speed_cap() in PCIe failed link retraining
PCI: Always lift 2.5GT/s restriction in PCIe failed link retraining
PCI: Remove MPS/MRRS Kconfig settings (CONFIG_PCIE_BUS_*)
- Don't reconfigure ASPM when entering low-power state; only do it
when returning back to D0 (Carlos Bilbao)
* pci/aspm:
PCI/ASPM: Don't reconfigure ASPM entering low-power state
Configured VLANs intermittently stop receiving traffic after a link
down/up cycle, e.g. when the network cable is unplugged and plugged back
in. VLAN filtering stays enabled but all VLAN-tagged frames are dropped
until a VLAN is added or removed again.
The LAN7801 datasheet (DS00002123E) states:
"A portion of the MAC operates on clocks generated by the Ethernet
PHY. During a PHY reset event, this portion of the MAC is designed to
not be taken out of reset until the PHY clocks are operational"
(section 8.10, MAC Reset Watchdog Timer)
"After a reset event, the RFE will automatically initialize the
contents of the VHF to 0h."
(section 7.1.4, VHF Organization)
Thus a link down/up cycle stops and restarts the PHY clock, resets the
PHY-clocked portion of the MAC, and the RFE clears its VLAN/DA hash
filter (VHF) memory. The VHF holds both the VLAN filter table and the
multicast hash table, but the driver never reprograms either from its
shadow copy once the link is back, so both stay empty.
Reprogram the VLAN filter and multicast hash tables on link up.
Reported-by: Sven Schuchmann <schuchmann@schleissheimer.de>
Closes: https://lore.kernel.org/netdev/BEZP281MB224501E38B30BFDC4BD3D364D9E32@BEZP281MB2245.DEUP281.PROD.OUTLOOK.COM/T/#u
Tested-by: Sven Schuchmann <schuchmann@schleissheimer.de>
Fixes: 55d7de9de6 ("Microchip's LAN7800 family USB 2/3 to 10/100/1000 Ethernet device driver")
Signed-off-by: Nicolai Buchwitz <nb@tipi-net.de>
Link: https://patch.msgid.link/20260622102911.484045-1-nb@tipi-net.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
During XDP enablement in veth, if xdp_rxq_info_reg() or
xdp_rxq_info_reg_mem_model() fails, the driver rolls back the changes.
However, the rollback loop:
for (i--; i >= start; i--) {
decrements the loop index 'i' before the first iteration. This
correctly skips unregistering the rxq for the failed index 'i' (as
registration failed or was already cleaned up), but it also
erroneously skips calling netif_napi_deli() for rq[i].xdp_napi.
Since netif_napi_add() was already called for index 'i', this leaves
a dangling napi_struct in the device's napi_list. When the veth
device is later destroyed, the freed queue memory (which contains the
leaked NAPI structure) can be reused.
The subsequent device teardown iterates the NAPI list and
corrupts the reallocated memory, leading to UAF.
Fix this by explicitly deleting the NAPI association for the failed
index 'i' before rolling back the successfully configured queues.
Fixes: b02e5a0ebb ("xsk: Propagate napi_id to XDP socket Rx path")
Reported-by: Guenter Roeck <groeck@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Björn Töpel <bjorn.topel@intel.com>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Cc: Michael S. Tsirkin <mst@redhat.com>
Cc: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Link: https://patch.msgid.link/20260622111825.88337-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
emac_xsk_xmit_zc() handles tx xmit for zero copy and gets called
inside napi context. User application wakes up the kernel while
initiating the transmit which triggers napi to start processing
the tx packets. The num_tx check inside emac_tx_complete_packets()
returns early if no packet transfer happen hindering the call
to emac_xsk_xmit_zc(). Remove this check to let application
wakeup initiate zero copy xmit traffic.
Add __netif_tx_lock() to ensure that the TX queue is protected
from concurrent access during the transmission of XDP frames.
This fixes netdev watchdog timeout for long runs.
Fixes: e2dc7bfd67 ("net: ti: icssg-prueth: Move common functions into a separate file")
Signed-off-by: Meghana Malladi <m-malladi@ti.com>
Link: https://patch.msgid.link/20260618100348.2209907-1-m-malladi@ti.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
pin_duration is converted from the user-provided period to SJA1105
clock ticks and is later passed as the cycle_time argument to
future_base_time().
Very small period values may become zero after the conversion,
which can lead to a division by zero in future_base_time().
Round zero pin_duration up to 1 tick so that the smallest unsupported
periods use the minimum non-zero hardware duration instead of passing
zero to future_base_time().
Fixes: 747e5eb31d ("net: dsa: sja1105: configure the PTP_CLK pin as EXT_TS or PER_OUT")
Signed-off-by: Aleksandrova Alyona <aga@itb.spb.ru>
Link: https://patch.msgid.link/20260618110508.53094-1-aga@itb.spb.ru
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Marek Szyprowski reported a deadlock during system resume when virtio_net
driver is used.
The deadlock occurs because netif_device_attach() is called while holding
dev->tx_global_lock (via netif_tx_lock_bh() in virtnet_restore_up()).
netif_device_attach() calls __netdev_watchdog_up(), which now also tries
to acquire dev->tx_global_lock to synchronize with dev_watchdog().
This recursive lock acquisition results in a deadlock.
Fix this by removing the tx_global_lock acquisition from netdev_watchdog_up().
The critical state (watchdog_timer and watchdog_ref_held) is already
protected by dev->watchdog_lock, which was introduced in the blamed commit.
Fixes: 8eed5519e4 ("net: watchdog: fix refcount tracking races")
Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
Closes: https://lore.kernel.org/netdev/a443376e-5187-4268-93b3-58047ef113a8@samsung.com/
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Marek Szyprowski <m.szyprowski@samsung.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Link: https://patch.msgid.link/20260622110108.69541-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull soundwire updates from Vinod Koul:
- Improvements in handling of soundwire groups
- Additional checks flagged by various tools
- Intel driver updates for ghost Realtek device handling in firmware
and adding devices to wake lists
* tag 'soundwire-7.2-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul/soundwire:
soundwire: dmi-quirks: Disable ghost Realtek devices
soundwire: only handle alert events when the peripheral is attached
soundwire: intel_ace2x: release bpt_stream when close it
soundwire: intel: Move suspend tracking from trigger to pm suspend
soundwire: intel_auxdevice: Add es9356 to wake_capable_list
soundwire: use krealloc_array to prevent integer overflow
soundwire: increase group->max_size after allocation
soundwire: fix bug in sdw_add_element_group_count found by syzkaller
soundwire: don't program SDW_SCP_BUSCLOCK_SCALE on a unattached Peripheral
soundwire: validate DT compatible before parsing it
soundwire: intel_auxdevice: Add cs42l43b to wake_capable_list
soundwire: stream: sdw_stream_remove_slave(): Check stream is valid
xdp_master_redirect() dereferences the result of
netdev_master_upper_dev_get_rcu() without a NULL check, but that helper
returns NULL when the receiving device has no upper-master adjacency.
The reach guard only checks netif_is_bond_slave(). On bond slave release
bond_upper_dev_unlink() drops the upper-master adjacency before clearing
IFF_SLAVE, so an XDP_TX reaching xdp_master_redirect() in that window
still passes netif_is_bond_slave() while master is already NULL, and
faults on master->flags at offset 0xb0:
BUG: kernel NULL pointer dereference, address: 00000000000000b0
RIP: 0010:xdp_master_redirect (net/core/filter.c:4432)
Call Trace:
xdp_master_redirect (net/core/filter.c:4432)
bpf_prog_run_generic_xdp (include/net/xdp.h:700)
do_xdp_generic (net/core/dev.c:5608)
__netif_receive_skb_one_core (net/core/dev.c:6204)
process_backlog (net/core/dev.c:6319)
__napi_poll (net/core/dev.c:7729)
net_rx_action (net/core/dev.c:7792)
handle_softirqs (kernel/softirq.c:622)
__dev_queue_xmit (include/linux/bottom_half.h:33)
packet_sendmsg (net/packet/af_packet.c:3082)
__sys_sendto (net/socket.c:2252)
Kernel panic - not syncing: Fatal exception in interrupt
The missing check dates back to the original code; commit 1921f91298
("net, bpf: fix null-ptr-deref in xdp_master_redirect() for down master")
later added the master->flags read where the fault now lands but kept the
unconditional deref. Check master for NULL before use; a NULL master is
treated the same as one that is not up.
Fixes: 879af96ffd ("net, core: Add support for XDP redirection to slave device")
Reported-by: Weiming Shi <bestswngs@gmail.com>
Signed-off-by: Xiang Mei <xmei5@asu.edu>
Reviewed-by: Jiayuan Chen <jiayuan.chen@linux.dev>
Link: https://patch.msgid.link/20260620201531.180123-1-xmei5@asu.edu
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c contains 7 (for
now) functions that use STATIC_IFN_KUNIT or INLINE_IFN_KUNIT macros for
function qualifiers (static or not, inline or not).
These cause parse warnings from kernel-doc:
Invalid C declaration: Expected identifier in nested name, got keyword:
struct [error at 29]
STATIC_IFN_KUNIT const struct drm_color_lut * __extract_blob_lut (const
struct drm_property_blob *blob, uint32_t *size)
Handle these in kernel-doc to prevent multiple warnings.
Fixes: 647d1fd046 ("drm/amd/display: Add KUnit test for color helpers")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Message-ID: <20260612234458.1084156-1-rdunlap@infradead.org>
Tushar Vyavahare says:
====================
selftests/xsk: stabilize timeout test behavior
This series improves AF_XDP selftests by making timeout handling
explicit and fixing sources of non-determinism in xsk timeout tests.
Patch 1 introduces test_spec::poll_tmout and removes implicit
dependence on RX UMEM setup state for timeout behavior.
Patch 2 fixes thread harness sequencing by attaching XDP programs
before worker startup, removing signal-based termination, and using
barrier synchronization only for dual-thread runs.
Patch 3 restores shared_umem after POLL_TXQ_FULL so test-local
configuration does not leak into subsequent cases on shared-netdev
runs.
Together these changes make timeout handling easier to follow and
improve selftest stability, especially on real NIC runs.
====================
Link: https://patch.msgid.link/20260616154955.1492560-1-tushar.vyavahare@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
POLL_TXQ_FULL temporarily disables shared_umem on TX to exercise the
TX timeout path in isolation.
With shared_umem enabled, TX setup expects RX UMEM to be initialized
first and fails with: "RX UMEM is not initialized before shared-UMEM TX
setup".
Save and restore shared_umem around POLL_TXQ_FULL execution, and restore
it on both success and pkt_stream_replace() failure paths.
Also add an in-code comment explaining why shared_umem is temporarily
disabled in this test.
This keeps timeout setup local and prevents cross-test state leakage.
Signed-off-by: Tushar Vyavahare <tushar.vyavahare@intel.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://patch.msgid.link/20260616154955.1492560-4-tushar.vyavahare@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Prevent workers from running before XDP program attachment completes.
The previous ordering allowed races between worker startup and setup.
Attach XDP programs before entering traffic validation.
Remove SIGUSR1-based worker termination and use pthread_join() for
thread shutdown so blocking syscalls are not interrupted.
Use barriers only for dual-thread runs so participants match and
teardown ordering stays deterministic.
This removes setup/startup races and stabilizes harness sequencing.
Signed-off-by: Tushar Vyavahare <tushar.vyavahare@intel.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://patch.msgid.link/20260616154955.1492560-3-tushar.vyavahare@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Stop inferring timeout behavior from RX UMEM initialization state.
That ties timeout semantics to setup internals and obscures intent.
Use test_spec::poll_tmout as the explicit timeout-mode selector in
TX and RX paths.
In RX, treat poll timeout as expected only in timeout mode.
In TX, let send_pkts() own loop completion in non-timeout mode
and use __send_pkts() only for progress and timeout detection.
This makes timeout logic explicit and keeps control flow predictable.
Signed-off-by: Magnus Karlsson <magnus.karlsson@intel.com>
Signed-off-by: Tushar Vyavahare <tushar.vyavahare@intel.com>
Reviewed-by: Jason Xing <kerneljasonxing@gmail.com>
Acked-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Tested-by: Maciej Fijalkowski <maciej.fijalkowski@intel.com>
Link: https://patch.msgid.link/20260616154955.1492560-2-tushar.vyavahare@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add support for DECLARE_PER_CPU() as a var (variable) as used in
<linux/netfilter/x_tables.h>.
Warning: include/linux/netfilter/x_tables.h:345 function parameter 'seqcount_t' not described in 'DECLARE_PER_CPU'
Warning: include/linux/netfilter/x_tables.h:345 function parameter 'xt_recseq' not described in 'DECLARE_PER_CPU'
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Message-ID: <20260614052452.1557987-1-rdunlap@infradead.org>
Pull sched_ext tree reorg from Tejun Heo:
"Pure source reorganization with no functional change:
- the kernel/sched/ext* files move into a new kernel/sched/ext/
subdirectory
- the headers and sources are made self-contained so editor tooling
can parse each file on its own"
* tag 'sched_ext-for-7.2-1' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext:
sched_ext: Move shared helpers from ext.c into internal.h and cid.h
sched_ext: Make kernel/sched/ext/ sources self-contained for clangd
sched_ext: Move sources under kernel/sched/ext/
The UIO howto still documents an mmap callback in struct uio_info.
That field was replaced by mmap_prepare, which takes a struct
vm_area_desc.
A UIO driver following the current howto no longer builds because
struct uio_info has no mmap member. Update the documented callback
signature and matching text to match the current API.
Fixes: 933f05f58a ("uio: replace deprecated mmap hook with mmap_prepare in uio_info")
Signed-off-by: Doehyun Baek <doehyunbaek@gmail.com>
Reviewed-by: Lorenzo Stoakes <ljs@kernel.org>
Acked-by: Randy Dunlap <rdunlap@infradead.org>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Message-ID: <20260622181821.1195257-1-doehyunbaek@gmail.com>
Add support for __SYSFS_FUNCTION_ALTERNATIVE() to create a union of its
members (as though CONFIG_CFI is unset).
Fixes these docs build warnings:
WARNING: include/linux/device.h:117 Invalid param: __SYSFS_FUNCTION_ALTERNATIVE( ssize_t (*show)(struct device *dev, struct device_attribute *attr, char *buf)
WARNING: include/linux/device.h:117 struct member '__SYSFS_FUNCTION_ALTERNATIVE( ssize_t (*show' not described in 'device_attribute'
WARNING: include/linux/device.h:117 Invalid param: __SYSFS_FUNCTION_ALTERNATIVE( ssize_t (*store)(struct device *dev, struct device_attribute *attr, const char *buf, size_t count)
WARNING: include/linux/device.h:117 struct member '__SYSFS_FUNCTION_ALTERNATIVE( ssize_t (*store' not described in 'device_attribute'
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reviewed-by: Thomas Weißschuh <linux@weissschuh.net>
Signed-off-by: Jonathan Corbet <corbet@lwn.net>
Message-ID: <20260623190006.406571-1-rdunlap@infradead.org>
pci_write_rom() controls access to the ROM content through the
corresponding sysfs attribute, and treats the input as a request to
disable only when it matches the string "0\n" exactly:
if ((off == 0) && (*buf == '0') && (count == 2))
The count == 2 condition encodes the trailing newline that echo(1) appends.
This was found when userspace wrote "0" without a trailing newline aiming
to disable access, which failed to match the condition above and enabled
access instead. For example:
$ echo 0 > rom # "0\n", count 2, access disabled
$ echo -n 0 > rom # "0", count 1, access enabled
$ echo > rom # "", count 1, access enabled (likely not desirable)
Parse the input with kstrtobool(), which handles common boolean inputs such
as "0", "1", "n", "y" or "off", "on", with or without a trailing newline,
so both of the above disable access, and update the now stale comment.
As a side effect, input that does not parse as a boolean is rejected with
-EINVAL rather than enabling access. The documented "0" and "1" continue
to work as before, and rejecting malformed input brings the attribute in
line with how sysfs attributes typically handle it.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://patch.msgid.link/20260612182448.552406-1-kwilczynski@kernel.org
Currently, __resource_resize_store() uses sysfs_remove_groups() and
sysfs_create_groups() on pci_dev_resource_attr_groups to tear down and
recreate the resourceN files after a BAR resize, so the updated BAR sizes
are visible in sysfs.
The resourceN files only exist on platforms that define HAVE_PCI_MMAP or
ARCH_GENERIC_PCI_MMAP_RESOURCE. On platforms that define neither,
pci_dev_resource_attr_groups is NULL and the sysfs_remove_groups() and
sysfs_create_groups() calls in __resource_resize_store() become no-ops.
Resizable BAR (ReBAR) is a PCI Express Extended Capability
(PCI_EXT_CAP_ID_REBAR) that requires PCIe extended config space. Every
PCIe-capable architecture defines HAVE_PCI_MMAP or
ARCH_GENERIC_PCI_MMAP_RESOURCE (via arch headers or the asm-generic/pci.h
fallback). Architectures without either only support conventional PCI and
cannot have any ReBAR-capable devices.
Move the resize show and store helpers, the per-BAR attribute definitions,
and the attribute group behind the existing #ifdef HAVE_PCI_MMAP ||
ARCH_GENERIC_PCI_MMAP_RESOURCE guard, and fold the group reference in
pci_dev_groups[] into the existing #if block.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Link: https://patch.msgid.link/20260508043543.217179-25-kwilczynski@kernel.org
Currently, pci_create_legacy_files() and pci_remove_legacy_files() are
no-op stubs. With legacy attributes now handled by static groups
registered via pcibus_groups[], no call site needs them.
Remove both functions, their declarations, and the call sites in
pci_register_host_bridge(), pci_alloc_child_bus(), and pci_remove_bus().
Remove the pci_sysfs_init() late_initcall and sysfs_initialized. The
late_initcall originally existed to create all the dynamic PCI sysfs files,
but with both resource and legacy attributes now handled by static groups,
it is no longer needed.
Remove the legacy_io and legacy_mem fields from struct pci_bus which were
used to track the dynamically allocated legacy attributes.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Link: https://patch.msgid.link/20260508043543.217179-24-kwilczynski@kernel.org
Currently, legacy_io and legacy_mem are dynamically allocated and created
by pci_create_legacy_files(), with pci_adjust_legacy_attr() updating the
attributes at runtime on Alpha to rename them and shift the size for sparse
addressing.
Convert to four static const attributes (legacy_io, legacy_io_sparse,
legacy_mem, legacy_mem_sparse) with .is_bin_visible() callbacks that use
pci_legacy_has_sparse() to select the appropriate variant per bus. The
sizes are compile-time constants and .size is set directly on each
attribute.
Register the groups in pcibus_groups[] under a HAVE_PCI_LEGACY guard so the
driver model handles creation and removal automatically.
Stub out pci_create_legacy_files() and pci_remove_legacy_files() as the
dynamic creation is no longer needed. Remove the __weak
pci_adjust_legacy_attr(), Alpha's override, and its declaration from both
Alpha and PowerPC asm/pci.h headers.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Link: https://patch.msgid.link/20260508043543.217179-23-kwilczynski@kernel.org
Currently, Alpha's sparse/dense legacy attribute handling is done via
pci_adjust_legacy_attr(), which updates dynamically allocated attributes at
runtime. The upcoming conversion to static attributes needs a way to
determine sparse support at visibility check time.
Add a __weak pci_legacy_has_sparse() that returns false by default. Alpha
overrides it to check has_sparse() on the bus host controller.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Link: https://patch.msgid.link/20260508043543.217179-22-kwilczynski@kernel.org
Currently, pci_mmap_legacy_page_range() reads the legacy resource size from
bus->legacy_mem->size or bus->legacy_io->size. This couples the mmap
bounds check to the struct pci_bus fields that will be removed when legacy
attributes are converted to static definitions.
Compute the size directly using PCI_LEGACY_MEM_SIZE (0x100000) and
PCI_LEGACY_IO_SIZE (0xffff) macros, and shift by 5 bits for sparse systems.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Magnus Lindholm <linmag7@gmail.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Acked-by: Magnus Lindholm <linmag7@gmail.com>
Link: https://patch.msgid.link/20260508043543.217179-21-kwilczynski@kernel.org
Currently, pci_create_sysfs_dev_files() and pci_remove_sysfs_dev_files()
are no-op stubs. With both the generic and Alpha resource files now
handled by static attribute groups, no platform needs dynamic per-device
sysfs file creation.
Remove both functions, their declarations, and the call sites in
pci_bus_add_device() and pci_stop_dev().
Remove __weak pci_create_resource_files() and pci_remove_resource_files()
stubs and their declarations in pci.h, as no architecture overrides them
anymore.
Remove the res_attr[] and res_attr_wc[] fields from struct pci_dev which
were used to track dynamically allocated resource attributes.
Finally, simplify pci_sysfs_init() to only handle legacy file creation
under HAVE_PCI_LEGACY, removing the per-device loop and the
HAVE_PCI_SYSFS_INIT helper added earlier.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Link: https://patch.msgid.link/20260508043543.217179-19-kwilczynski@kernel.org
Previously, Alpha's PCI resource files (resourceN, resourceN_sparse,
resourceN_dense) were dynamically created by pci_create_resource_files(),
which overrides the generic __weak implementation. The previous code
allocated bin_attributes at runtime and managed them via the res_attr[] and
res_attr_wc[] fields in struct pci_dev.
Convert to static const attributes with three attribute groups (plain,
sparse, dense), each with an .is_bin_visible() callback that checks
resource length, has_sparse(), and sparse_mem_mmap_fits(). A .bin_size()
callback provides the resource size to the kernfs node, with the sparse
variant shifting by 5 bits for byte-level addressing.
Register the groups via ARCH_PCI_DEV_GROUPS so the driver model handles
creation and removal automatically.
Use the new pci_resource_is_mem() helper for the type check, replacing the
open-coded bitwise flag test.
Finally, remove pci_create_resource_files(), pci_remove_resource_files(),
pci_create_attr(), and pci_create_one_attr() which are no longer needed.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Magnus Lindholm <linmag7@gmail.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Acked-by: Magnus Lindholm <linmag7@gmail.com>
Link: https://patch.msgid.link/20260508043543.217179-18-kwilczynski@kernel.org
Add macros for declaring static binary attributes for Alpha's PCI resource
files:
- pci_dev_resource_attr(), for dense/BWX systems (mmap dense)
- pci_dev_resource_sparse_attr(), for sparse systems (mmap sparse)
- pci_dev_resource_dense_attr(), for dense companion files (mmap dense)
Each macro creates a const bin_attribute with the BAR index stored in the
.private property and the appropriate .mmap() callback.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Magnus Lindholm <linmag7@gmail.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Acked-by: Magnus Lindholm <linmag7@gmail.com>
Link: https://patch.msgid.link/20260508043543.217179-17-kwilczynski@kernel.org
Currently, Alpha's pci_create_one_attr() stores a resource pointer in
attr->private, and pci_mmap_resource() loops through all BARs to find
the matching index.
Store the BAR index directly in attr->private and retrieve the resource via
pci_resource_n(). This eliminates the loop and aligns with the convention
used by the generic PCI sysfs code.
The PCI core change was first added in the commit dca40b186b ("PCI: Use
BAR index in sysfs attr->private instead of resource pointer").
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Magnus Lindholm <linmag7@gmail.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Reviewed-by: Ilpo Järvinen <ilpo.jarvinen@linux.intel.com>
Acked-by: Magnus Lindholm <linmag7@gmail.com>
Link: https://patch.msgid.link/20260508043543.217179-13-kwilczynski@kernel.org
Currently, pci_sysfs_init() and sysfs_initialized compile unconditionally,
even on platforms where static attribute groups handle all resource file
creation.
Place them behind a new HAVE_PCI_SYSFS_INIT macro, especially as the
late_initcall is only needed when:
- HAVE_PCI_LEGACY is set, to iterate buses and create legacy I/O and
memory files.
- Neither HAVE_PCI_MMAP nor ARCH_GENERIC_PCI_MMAP_RESOURCE is set, to
iterate devices and create resource files via the __weak
pci_create_resource_files() stub override (this is how the Alpha
architecture handles this currently).
On most systems both conditions are false and the entire late_initcall
compiles away.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Link: https://patch.msgid.link/20260508043543.217179-11-kwilczynski@kernel.org
On platforms with HAVE_PCI_MMAP or ARCH_GENERIC_PCI_MMAP_RESOURCE, resource
files are now handled by static attribute groups registered via
pci_dev_groups[].
Stub out the pci_create_sysfs_dev_files() and pci_remove_sysfs_dev_files(),
as the dynamic resource file creation is no longer needed.
Also, simplify pci_sysfs_init() on these platforms to only iterate buses
for legacy attributes creation, skipping the per-device loop.
Move the __weak stubs for pci_create_resource_files() and
pci_remove_resource_files() into the #else branch since only platforms
without HAVE_PCI_MMAP (such as Alpha architecture) still need them. Guard
the res_attr[] and res_attr_wc[] fields in struct pci_dev the same way.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Link: https://patch.msgid.link/20260508043543.217179-10-kwilczynski@kernel.org
Add a pci_warn() to __resource_resize_store(), so that BAR resize failures
are visible to the user, which can help troubleshoot any potential resource
resize issues.
While at it, rename the resource_resize_is_visible() to
resource_resize_attr_is_visible() along with the corresponding group
variable to align with the naming convention used by the resource attribute
groups.
Also, change the order of pci_dev_groups[] such that the resize group is
now located alongside the other resource groups.
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Link: https://patch.msgid.link/20260508043543.217179-9-kwilczynski@kernel.org
Currently, the PCI resource files (resourceN, resourceN_wc) are dynamically
created by pci_create_sysfs_dev_files(), called from both
pci_bus_add_device() and the pci_sysfs_init() late_initcall, with only a
sysfs_initialized flag for synchronisation. This has caused warnings and
boot panics when both paths race on the same device, e.g.:
sysfs: cannot create duplicate filename '/devices/pci0000:3c/0000:3c:01.0/0000:3e:00.2/resource2'
This is especially likely on Devicetree-based platforms, where the PCI host
controllers are platform drivers that probe via the driver model, which can
happen during or after the late_initcall. As such, pci_bus_add_device()
and pci_sysfs_init() are more likely to overlap.
Convert to static const attributes with three attribute groups (I/O, UC,
WC), each with an .is_bin_visible() callback that checks resource flags,
BAR length, and non_mappable_bars. A .bin_size() callback provides
pci_resource_len() to the kernfs node for correct stat and lseek behaviour.
As part of this conversion:
- Rename pci_read_resource_io() and pci_write_resource_io() to
pci_read_resource() and pci_write_resource() since the callbacks are no
longer I/O-specific in the static attribute context.
- Update __resource_resize_store() to use sysfs_create_groups() and
sysfs_remove_groups(), which re-evaluates visibility and runs the
.bin_size() callback for the static resource attribute groups.
- Remove pci_create_resource_files(), pci_remove_resource_files(), and
pci_create_attr() which are no longer needed.
- Move the __weak stubs outside the #if guard so they remain available
for callers converted in subsequent commits.
Platforms that do not define the HAVE_PCI_MMAP macro or the
ARCH_GENERIC_PCI_MMAP_RESOURCE macro, such as Alpha architecture,
continue using their platform-specific resource file creation.
For reference, the dynamic creation dates back to the pre-Git era:
https://git.kernel.org/pub/scm/linux/kernel/git/tglx/history.git/commit/drivers/pci/pci-sysfs.c?id=42298be0eeb5ae98453b3374c36161b05a46c5dc
The write-combine support was added in commit 45aec1ae72 ("x86: PAT
export resource_wc in pci sysfs").
Many other reports mentioned in the cover letter (first Link: below).
Link: https://lore.kernel.org/r/20260508043543.217179-1-kwilczynski@kernel.org/
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=215515
Closes: https://github.com/openwrt/openwrt/issues/17143
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Tested-by: Shivaprasad G Bhat <sbhat@linux.ibm.com>
Link: https://patch.msgid.link/20260508043543.217179-8-kwilczynski@kernel.org
pci_proc_attach_device() creates procfs entries for PCI devices and is
called from pci_bus_add_device(). It lazily creates the per-bus procfs
directory (bus->procdir) via proc_mkdir() on first use, and returns early
if proc_initialized is not yet set.
On x86 with ACPI, PCI enumeration occurs at subsys_initcall, before
pci_proc_init() sets proc_initialized at device_initcall. The
for_each_pci_dev() loop in pci_proc_init() then creates procfs entries for
these already-enumerated devices, but runs without holding
pci_rescan_remove_lock.
On ARM64 with devicetree, PCI host bridges probe at device_initcall. With
async probing enabled, pci_bus_add_device() can run concurrently with
pci_proc_init(), and both may call pci_proc_attach_device() for the same
device or for different devices on the same bus. As pci_host_probe() holds
pci_rescan_remove_lock while pci_proc_init() does not, there is no
serialisation between the two paths.
When two threads concurrently call pci_proc_attach_device() for devices on
the same bus, both observe bus->procdir as NULL and both call proc_mkdir().
The proc filesystem serialises directory creation internally, so only one
caller succeeds. The other results in a warning like:
proc_dir_entry '000c:00/00.0' already registered
The caller receives NULL (duplicate entry) and unconditionally stores it to
bus->procdir, corrupting the valid pointer set by the first caller.
Serialise access to proc_initialized, proc_bus_pci_dir, bus->procdir and
dev->procent with a new mutex local to drivers/pci/proc.c, and store the
created entries to bus->procdir and dev->procent only on success, so a
failed creation can never overwrite a valid pointer.
Additionally, wrap the for_each_pci_dev() loop in pci_proc_init() with
pci_lock_rescan_remove() to serialise against concurrent PCI bus
operations, add an early return in pci_proc_attach_device() when
dev->procent is already set to make the function idempotent, and clear
bus->procdir in pci_proc_detach_bus() to prevent use of a dangling pointer
after proc_remove().
Reported-by: Shuan He <heshuan@bytedance.com>
Closes: https://lore.kernel.org/linux-pci/20250702155112.40124-2-heshuan@bytedance.com/
Signed-off-by: Krzysztof Wilczyński <kwilczynski@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://lore.kernel.org/r/20260611150543.511422-1-kwilczynski@kernel.org
Replace the unconditional msleep(100) with the common helper
pci_host_common_link_train_delay(). The helper only waits when
max_link_speed > 2, as required by PCIe r6.0 sec 6.6.1.
This avoids unnecessary delay for Gen1/Gen2 links while retaining
the mandatory 100 ms for higher speeds.
Signed-off-by: Hans Zhang <18255117159@163.com>
Signed-off-by: Manivannan Sadhasivam <mani@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://patch.msgid.link/20260518004246.1384532-8-18255117159@163.com
The MediaTek Gen3 PCIe host driver lacks the required 100 ms delay after
link training completes for speeds > 5.0 GT/s, as specified in PCIe r6.0
sec 6.6.1.
The driver already stores max_link_speed (from the device tree). After
mtk_pcie_startup_port() successfully brings up the link, call
pci_host_common_link_train_delay() to comply with the specification.
Signed-off-by: Hans Zhang <18255117159@163.com>
Signed-off-by: Manivannan Sadhasivam <mani@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://patch.msgid.link/20260518004246.1384532-7-18255117159@163.com
The Aardvark PCIe controller driver waits for the link to come up but
does not implement the mandatory 100 ms delay after link training
completes for speeds greater than 5.0 GT/s (PCIe r6.0 sec 6.6.1).
The driver already maintains a 'link_gen' field that holds the negotiated
link speed. Use it together with pci_host_common_link_train_delay() to
insert the required delay immediately after confirming that the link
is up.
Signed-off-by: Hans Zhang <18255117159@163.com>
Signed-off-by: Manivannan Sadhasivam <mani@kernel.org>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Link: https://patch.msgid.link/20260518004246.1384532-6-18255117159@163.com
Bjorn Helgaas
Nicolai Buchwitz
Eric Dumazet
Meghana Malladi
Aleksandrova Alyona
Linus Torvalds
Declan Wale
Xiang Mei
Randy Dunlap
Jakub Kicinski
Tushar Vyavahare
Matthew Wilcox (Oracle)
Zenghui Yu
Yudistira Putra
Doehyun Baek
David Hildenbrand (Arm)
Krzysztof Wilczyński
Hans Zhang