| Age | Commit message (Collapse) | Author |
|---|
|
git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux
Pull RISC-V fixes from Palmer Dabbelt:
- The compressed half-word misaligned access instructions (c.lhu, c.lh,
and c.sh) from the Zcb extension are now properly emulated
- A series of fixes to properly emulate permissions while handling
userspace misaligned accesses
- A pair of fixes for PR_GET_TAGGED_ADDR_CTRL to avoid accessing the
envcfg CSR on systems that don't support that CSR, and to report
those failures up to userspace
- The .rela.dyn section is no longer stripped from vmlinux, as it is
necessary to relocate the kernel under some conditions (including
kexec)
* tag 'riscv-for-linus-6.15-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/riscv/linux:
riscv: Disallow PR_GET_TAGGED_ADDR_CTRL without Supm
scripts: Do not strip .rela.dyn section
riscv: Fix kernel crash due to PR_SET_TAGGED_ADDR_CTRL
riscv: misaligned: use get_user() instead of __get_user()
riscv: misaligned: enable IRQs while handling misaligned accesses
riscv: misaligned: factorize trap handling
riscv: misaligned: Add handling for ZCB instructions
|
|
Pull block fixes from Jens Axboe:
- Fix for a regression in this series for loop and read/write iterator
handling
- zone append block update tweak
- remove a broken IO priority test
- NVMe pull request via Christoph:
- unblock ctrl state transition for firmware update (Daniel
Wagner)
* tag 'block-6.15-20250509' of git://git.kernel.dk/linux:
block: remove test of incorrect io priority level
nvme: unblock ctrl state transition for firmware update
block: only update request sector if needed
loop: Add sanity check for read/write_iter
|
|
Pull io_uring fixes from Jens Axboe:
- Fix for linked timeouts arming and firing wrt prep and issue of the
request being managed by the linked timeout
- Fix for a CQE ordering issue between requests with multishot and
using the same buffer group. This is a dumbed down version for this
release and for stable, it'll get improved for v6.16
- Tweak the SQPOLL submit batch size. A previous commit made SQPOLL
manage its own task_work and chose a tiny batch size, bump it from 8
to 32 to fix a performance regression due to that
* tag 'io_uring-6.15-20250509' of git://git.kernel.dk/linux:
io_uring/sqpoll: Increase task_work submission batch size
io_uring: ensure deferred completions are flushed for multishot
io_uring: always arm linked timeouts prior to issue
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/modules/linux
Pull modules fix from Petr Pavlu:
"A single fix to prevent use of an uninitialized completion pointer
when releasing a module_kobject in specific situations.
This addresses a latent bug exposed by commit f95bbfe18512 ("drivers:
base: handle module_kobject creation")"
* tag 'modules-6.15-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/modules/linux:
module: ensure that kobject_put() is safe for module type kobjects
|
|
tl;dr: There is a window in the mm switching code where the new CR3 is
set and the CPU should be getting TLB flushes for the new mm. But
should_flush_tlb() has a bug and suppresses the flush. Fix it by
widening the window where should_flush_tlb() sends an IPI.
Long Version:
=== History ===
There were a few things leading up to this.
First, updating mm_cpumask() was observed to be too expensive, so it was
made lazier. But being lazy caused too many unnecessary IPIs to CPUs
due to the now-lazy mm_cpumask(). So code was added to cull
mm_cpumask() periodically[2]. But that culling was a bit too aggressive
and skipped sending TLB flushes to CPUs that need them. So here we are
again.
=== Problem ===
The too-aggressive code in should_flush_tlb() strikes in this window:
// Turn on IPIs for this CPU/mm combination, but only
// if should_flush_tlb() agrees:
cpumask_set_cpu(cpu, mm_cpumask(next));
next_tlb_gen = atomic64_read(&next->context.tlb_gen);
choose_new_asid(next, next_tlb_gen, &new_asid, &need_flush);
load_new_mm_cr3(need_flush);
// ^ After 'need_flush' is set to false, IPIs *MUST*
// be sent to this CPU and not be ignored.
this_cpu_write(cpu_tlbstate.loaded_mm, next);
// ^ Not until this point does should_flush_tlb()
// become true!
should_flush_tlb() will suppress TLB flushes between load_new_mm_cr3()
and writing to 'loaded_mm', which is a window where they should not be
suppressed. Whoops.
=== Solution ===
Thankfully, the fuzzy "just about to write CR3" window is already marked
with loaded_mm==LOADED_MM_SWITCHING. Simply checking for that state in
should_flush_tlb() is sufficient to ensure that the CPU is targeted with
an IPI.
This will cause more TLB flush IPIs. But the window is relatively small
and I do not expect this to cause any kind of measurable performance
impact.
Update the comment where LOADED_MM_SWITCHING is written since it grew
yet another user.
Peter Z also raised a concern that should_flush_tlb() might not observe
'loaded_mm' and 'is_lazy' in the same order that switch_mm_irqs_off()
writes them. Add a barrier to ensure that they are observed in the
order they are written.
Signed-off-by: Dave Hansen <dave.hansen@linux.intel.com>
Acked-by: Rik van Riel <riel@surriel.com>
Link: https://lore.kernel.org/oe-lkp/202411282207.6bd28eae-lkp@intel.com/ [1]
Fixes: 6db2526c1d69 ("x86/mm/tlb: Only trim the mm_cpumask once a second") [2]
Reported-by: Stephen Dolan <sdolan@janestreet.com>
Cc: stable@vger.kernel.org
Acked-by: Ingo Molnar <mingo@kernel.org>
Acked-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Our QA team reported a 10%-23%, throughput reduction on an io_uring
sqpoll testcase doing IO to a null_blk, that I traced back to a
reduction of the device submission queue depth utilization. It turns out
that, after commit af5d68f8892f ("io_uring/sqpoll: manage task_work
privately"), we capped the number of task_work entries that can be
completed from a single spin of sqpoll to only 8 entries, before the
sqpoll goes around to (potentially) sleep. While this cap doesn't drive
the submission side directly, it impacts the completion behavior, which
affects the number of IO queued by fio per sqpoll cycle on the
submission side, and io_uring ends up seeing less ios per sqpoll cycle.
As a result, block layer plugging is less effective, and we see more
time spent inside the block layer in profilings charts, and increased
submission latency measured by fio.
There are other places that have increased overhead once sqpoll sleeps
more often, such as the sqpoll utilization calculation. But, in this
microbenchmark, those were not representative enough in perf charts, and
their removal didn't yield measurable changes in throughput. The major
overhead comes from the fact we plug less, and less often, when submitting
to the block layer.
My benchmark is:
fio --ioengine=io_uring --direct=1 --iodepth=128 --runtime=300 --bs=4k \
--invalidate=1 --time_based --ramp_time=10 --group_reporting=1 \
--filename=/dev/nullb0 --name=RandomReads-direct-nullb-sqpoll-4k-1 \
--rw=randread --numjobs=1 --sqthread_poll
In one machine, tested on top of Linux 6.15-rc1, we have the following
baseline:
READ: bw=4994MiB/s (5236MB/s), 4994MiB/s-4994MiB/s (5236MB/s-5236MB/s), io=439GiB (471GB), run=90001-90001msec
With this patch:
READ: bw=5762MiB/s (6042MB/s), 5762MiB/s-5762MiB/s (6042MB/s-6042MB/s), io=506GiB (544GB), run=90001-90001msec
which is a 15% improvement in measured bandwidth. The average
submission latency is noticeably lowered too. As measured by
fio:
Baseline:
lat (usec): min=20, max=241, avg=99.81, stdev=3.38
Patched:
lat (usec): min=26, max=226, avg=86.48, stdev=4.82
If we look at blktrace, we can also see the plugging behavior is
improved. In the baseline, we end up limited to plugging 8 requests in
the block layer regardless of the device queue depth size, while after
patching we can drive more io, and we manage to utilize the full device
queue.
In the baseline, after a stabilization phase, an ordinary submission
looks like:
254,0 1 49942 0.016028795 5977 U N [iou-sqp-5976] 7
After patching, I see consistently more requests per unplug.
254,0 1 4996 0.001432872 3145 U N [iou-sqp-3144] 32
Ideally, the cap size would at least be the deep enough to fill the
device queue, but we can't predict that behavior, or assume all IO goes
to a single device, and thus can't guess the ideal batch size. We also
don't want to let the tw run unbounded, though I'm not sure it would
really be a problem. Instead, let's just give it a more sensible value
that will allow for more efficient batching. I've tested with different
cap values, and initially proposed to increase the cap to 1024. Jens
argued it is too big of a bump and I observed that, with 32, I'm no
longer able to observe this bottleneck in any of my machines.
Fixes: af5d68f8892f ("io_uring/sqpoll: manage task_work privately")
Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
Link: https://lore.kernel.org/r/20250508181203.3785544-1-krisman@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Pull bcachefs fixes from Kent Overstreet:
- Some fixes to help with filesystem analysis: ensure superblock
error count gets written if we go ERO, don't discard the journal
aggressively (so it's available for list_journal -a)
- Fix lost wakeup on arm causing us to get stuck when reading btree
nodes
- Fix fsck failing to exit on ctrl-c
- An additional fix for filesystems with misaligned bucket sizes: we
now ensure that allocations are properly aligned
- Setting background target but not promote target will now leave that
data cached on the foreground target, as it used to
- Revert a change to when we allocate the VFS superblock, this was done
for implementing blk_holder_ops but ended up not being needed, and
allocating a superblock and not setting SB_BORN while we do recovery
caused sync() calls and other things to hang
- Assorted fixes for harmless error messages that caused concern to
users
* tag 'bcachefs-2025-05-08' of git://evilpiepirate.org/bcachefs:
bcachefs: Don't aggressively discard the journal
bcachefs: Ensure superblock gets written when we go ERO
bcachefs: Filter out harmless EROFS error messages
bcachefs: journal_shutdown is EROFS, not EIO
bcachefs: Call bch2_fs_start before getting vfs superblock
bcachefs: fix hung task timeout in journal read
bcachefs: Add missing barriers before wake_up_bit()
bcachefs: Ensure proper write alignment
bcachefs: Improve want_cached_ptr()
bcachefs: thread_with_stdio: fix spinning instead of exiting
|
|
Pull vfio fix from Alex Williamson:
- Fix an issue in vfio-pci huge_fault handling by aligning faults to
the order, resulting in deterministic use of huge pages. This
avoids a race where simultaneous aligned and unaligned faults to
the same PMD can result in a VM_FAULT_OOM and subsequent VM crash.
(Alex Williamson)
* tag 'vfio-v6.15-rc6' of https://github.com/awilliam/linux-vfio:
vfio/pci: Align huge faults to order
|
|
ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/alexghiti/linux into fixes
riscv fixes for 6.15-rc6
- A fix to handle compressed halfword load/store instructions misaligned accesses
- A fix to allow user memory access while handling a misaligned access
- 2 fixes to return an error if the pointer masking extension is not implemented on the platform but userspace still tries to access it, which caused oops on some early platforms
- A fix to prevent the stripping of .rela.dyn so that a vmlinux loaded by kexec can successfully boot
* tag 'riscv-fixes-6.15-rc6' of ssh://gitolite.kernel.org/pub/scm/linux/kernel/git/alexghiti/linux:
riscv: Disallow PR_GET_TAGGED_ADDR_CTRL without Supm
scripts: Do not strip .rela.dyn section
riscv: Fix kernel crash due to PR_SET_TAGGED_ADDR_CTRL
riscv: misaligned: use get_user() instead of __get_user()
riscv: misaligned: enable IRQs while handling misaligned accesses
riscv: misaligned: factorize trap handling
riscv: misaligned: Add handling for ZCB instructions
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Pull networking fixes from Paolo Abeni:
"Including fixes from CAN, WiFi and netfilter.
We have still a comple of regressions open due to the recent
drivers locking refactor. The patches are in-flight, but not
ready yet.
Current release - regressions:
- core: lock netdevices during dev_shutdown
- sch_htb: make htb_deactivate() idempotent
- eth: virtio-net: don't re-enable refill work too early
Current release - new code bugs:
- eth: icssg-prueth: fix kernel panic during concurrent Tx queue
access
Previous releases - regressions:
- gre: fix again IPv6 link-local address generation.
- eth: b53: fix learning on VLAN unaware bridges
Previous releases - always broken:
- wifi: fix out-of-bounds access during multi-link element
defragmentation
- can:
- initialize spin lock on device probe
- fix order of unregistration calls
- openvswitch: fix unsafe attribute parsing in output_userspace()
- eth:
- virtio-net: fix total qstat values
- mtk_eth_soc: reset all TX queues on DMA free
- fbnic: firmware IPC mailbox fixes"
* tag 'net-6.15-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (55 commits)
virtio-net: fix total qstat values
net: export a helper for adding up queue stats
fbnic: Do not allow mailbox to toggle to ready outside fbnic_mbx_poll_tx_ready
fbnic: Pull fbnic_fw_xmit_cap_msg use out of interrupt context
fbnic: Improve responsiveness of fbnic_mbx_poll_tx_ready
fbnic: Cleanup handling of completions
fbnic: Actually flush_tx instead of stalling out
fbnic: Add additional handling of IRQs
fbnic: Gate AXI read/write enabling on FW mailbox
fbnic: Fix initialization of mailbox descriptor rings
net: dsa: b53: do not set learning and unicast/multicast on up
net: dsa: b53: fix learning on VLAN unaware bridges
net: dsa: b53: fix toggling vlan_filtering
net: dsa: b53: do not program vlans when vlan filtering is off
net: dsa: b53: do not allow to configure VLAN 0
net: dsa: b53: always rejoin default untagged VLAN on bridge leave
net: dsa: b53: fix VLAN ID for untagged vlan on bridge leave
net: dsa: b53: fix flushing old pvid VLAN on pvid change
net: dsa: b53: fix clearing PVID of a port
net: dsa: b53: keep CPU port always tagged again
...
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Pull s390 fixes from Heiko Carstens:
- Fix potential use-after-free bug and missing error handling in PCI
code
- Fix dcssblk build error
- Fix last breaking event handling in case of stack corruption to allow
for better error reporting
- Update defconfigs
* tag 's390-6.15-4' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
s390/pci: Fix duplicate pci_dev_put() in disable_slot() when PF has child VFs
s390/pci: Fix missing check for zpci_create_device() error return
s390: Update defconfigs
s390/dcssblk: Fix build error with CONFIG_DAX=m and CONFIG_DCSSBLK=y
s390/entry: Fix last breaking event handling in case of stack corruption
s390/configs: Enable options required for TC flow offload
s390/configs: Enable VDPA on Nvidia ConnectX-6 network card
|
|
Pull smb server fixes from Steve French:
- Fix UAF closing file table (e.g. in tree disconnect)
- Fix potential out of bounds write
- Fix potential memory leak parsing lease state in open
- Fix oops in rename with empty target
* tag 'v6.15-rc5-ksmbd-server-fixes' of git://git.samba.org/ksmbd:
ksmbd: Fix UAF in __close_file_table_ids
ksmbd: prevent out-of-bounds stream writes by validating *pos
ksmbd: fix memory leak in parse_lease_state()
ksmbd: prevent rename with empty string
|
|
Pull NVMe fix from Christoph:
"nvme fixes for linux 6.15
- unblock ctrl state transition for firmware update (Daniel Wagner)"
* tag 'nvme-6.15-2025-05-08' of git://git.infradead.org/nvme:
nvme: unblock ctrl state transition for firmware update
|
|
Ever since commit eca2040972b4("scsi: block: ioprio: Clean up interface
definition"), the macro IOPRIO_PRIO_LEVEL() will mask the level value to
something between 0 and 7 so necessarily, level will always be lower than
IOPRIO_NR_LEVELS(8).
Remove this obsolete check.
Reported-by: Kexin Wei <ys.weikexin@h3c.com>
Cc: Damien Le Moal <dlemoal@kernel.org>
Signed-off-by: Aaron Lu <ziqianlu@bytedance.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20250508083018.GA769554@bytedance
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
When the prctl() interface for pointer masking was added, it did not
check that the pointer masking ISA extension was supported, only the
individual submodes. Userspace could still attempt to disable pointer
masking and query the pointer masking state. commit 81de1afb2dd1
("riscv: Fix kernel crash due to PR_SET_TAGGED_ADDR_CTRL") disallowed
the former, as the senvcfg write could crash on older systems.
PR_GET_TAGGED_ADDR_CTRL state does not crash, because it reads only
kernel-internal state and not senvcfg, but it should still be disallowed
for consistency.
Fixes: 09d6775f503b ("riscv: Add support for userspace pointer masking")
Signed-off-by: Samuel Holland <samuel.holland@sifive.com>
Reviewed-by: Nam Cao <namcao@linutronix.de>
Link: https://lore.kernel.org/r/20250507145230.2272871-1-samuel.holland@sifive.com
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
|
|
The .rela.dyn section contains runtime relocations and is only emitted
for a relocatable kernel.
riscv uses this section to relocate the kernel at runtime but that section
is stripped from vmlinux. That prevents kexec to successfully load vmlinux
since it does not contain the relocations info needed.
Fixes: 559d1e45a16d ("riscv: Use --emit-relocs in order to move .rela.dyn in init")
Tested-by: Björn Töpel <bjorn@rivosinc.com>
Reviewed-by: Björn Töpel <bjorn@rivosinc.com>
Acked-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20250408072851.90275-1-alexghiti@rivosinc.com
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
|
|
When userspace does PR_SET_TAGGED_ADDR_CTRL, but Supm extension is not
available, the kernel crashes:
Oops - illegal instruction [#1]
[snip]
epc : set_tagged_addr_ctrl+0x112/0x15a
ra : set_tagged_addr_ctrl+0x74/0x15a
epc : ffffffff80011ace ra : ffffffff80011a30 sp : ffffffc60039be10
[snip]
status: 0000000200000120 badaddr: 0000000010a79073 cause: 0000000000000002
set_tagged_addr_ctrl+0x112/0x15a
__riscv_sys_prctl+0x352/0x73c
do_trap_ecall_u+0x17c/0x20c
andle_exception+0x150/0x15c
Fix it by checking if Supm is available.
Fixes: 09d6775f503b ("riscv: Add support for userspace pointer masking")
Signed-off-by: Nam Cao <namcao@linutronix.de>
Cc: stable@vger.kernel.org
Reviewed-by: Samuel Holland <samuel.holland@sifive.com>
Link: https://lore.kernel.org/r/20250504101920.3393053-1-namcao@linutronix.de
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
|
|
Now that we can safely handle user memory accesses while in the
misaligned access handlers, use get_user() instead of __get_user() to
have user memory access checks.
Signed-off-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20250422162324.956065-4-cleger@rivosinc.com
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
|
|
We can safely reenable IRQs if coming from userspace. This allows to
access user memory that could potentially trigger a page fault.
Fixes: b686ecdeacf6 ("riscv: misaligned: Restrict user access to kernel memory")
Signed-off-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20250422162324.956065-3-cleger@rivosinc.com
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
|
|
Since both load/store and user/kernel should use almost the same path and
that we are going to add some code around that, factorize it.
Signed-off-by: Clément Léger <cleger@rivosinc.com>
Reviewed-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Link: https://lore.kernel.org/r/20250422162324.956065-2-cleger@rivosinc.com
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
|
|
Jakub Kicinski says:
====================
virtio-net: fix total qstat values
Another small fix discovered after we enabled virtio multi-queue
in netdev CI. The queue stat test fails:
# Exception| Exception: Qstats are lower, fetched later
not ok 3 stats.pkt_byte_sum
The queue stats from disabled queues are supposed to be reported
in the "base" stats.
====================
Link: https://patch.msgid.link/20250507003221.823267-1-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
NIPA tests report that the interface statistics reported
via qstat are lower than those reported via ip link.
Looks like this is because some tests flip the queue
count up and down, and we end up with some of the traffic
accounted on disabled queues.
Add up counters from disabled queues.
Fixes: d888f04c09bb ("virtio-net: support queue stat")
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Link: https://patch.msgid.link/20250507003221.823267-3-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Older drivers and drivers with lower queue counts often have a static
array of queues, rather than allocating structs for each queue on demand.
Add a helper for adding up qstats from a queue range. Expectation is
that driver will pass a queue range [netdev->real_num_*x_queues, MAX).
It was tempting to always use num_*x_queues as the end, but virtio
seems to clamp its queue count after allocating the netdev. And this
way we can trivaly reuse the helper for [0, real_..).
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Link: https://patch.msgid.link/20250507003221.823267-2-kuba@kernel.org
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Alexander Duyck says:
====================
fbnic: FW IPC Mailbox fixes
This series is meant to address a number of issues that have been found in
the FW IPC mailbox over the past several months.
The main issues addressed are:
1. Resolve a potential race between host and FW during initialization that
can cause the FW to only have the lower 32b of an address.
2. Block the FW from issuing DMA requests after we have closed the mailbox
and before we have started issuing requests on it.
3. Fix races in the IRQ handlers that can cause the IRQ to unmask itself if
it is being processed while we are trying to disable it.
4. Cleanup the Tx flush logic so that we actually lock down the Tx path
before we start flushing it instead of letting it free run while we are
shutting it down.
5. Fix several memory leaks that could occur if we failed initialization.
6. Cleanup the mailbox completion if we are flushing Tx since we are no
longer processing Rx.
7. Move several allocations out of a potential IRQ/atomic context.
There have been a few optimizations we also picked up since then. Rather
than split them out I just folded them into these diffs. They mostly
address minor issues such as how long it takes to initialize and/or fail so
I thought they could probably go in with the rest of the patches. They
consist of:
1. Do not sleep more than 20ms waiting on FW to respond as the 200ms value
likely originated from simulation/emulation testing.
2. Use jiffies to determine timeout instead of sleep * attempts for better
accuracy.
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
====================
Link: https://patch.msgid.link/174654659243.499179.11194817277075480209.stgit@ahduyck-xeon-server.home.arpa
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
We had originally thought to have the mailbox go to ready in the background
while we were doing other things. One issue with this though is that we
can't disable it by clearing the ready state without also blocking
interrupts or calls to mbx_poll as it will just pop back to life during an
interrupt.
In order to prevent that from happening we can pull the code for toggling
to ready out of the interrupt path and instead place it in the
fbnic_mbx_poll_tx_ready path so that it becomes the only spot where the
Rx/Tx can toggle to the ready state. By doing this we can prevent races
where we disable the DMA and/or free buffers only to have an interrupt fire
and undo what we have done.
Fixes: da3cde08209e ("eth: fbnic: Add FW communication mechanism")
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/174654722518.499179.11612865740376848478.stgit@ahduyck-xeon-server.home.arpa
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
This change pulls the call to fbnic_fw_xmit_cap_msg out of
fbnic_mbx_init_desc_ring and instead places it in the polling function for
getting the Tx ready. Doing that we can avoid the potential issue with an
interrupt coming in later from the firmware that causes it to get fired in
interrupt context.
Fixes: 20d2e88cc746 ("eth: fbnic: Add initial messaging to notify FW of our presence")
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/174654721876.499179.9839651602256668493.stgit@ahduyck-xeon-server.home.arpa
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
There were a couple different issues found in fbnic_mbx_poll_tx_ready.
Among them were the fact that we were sleeping much longer than we actually
needed to as the actual FW could respond in under 20ms. The other issue was
that we would just keep polling the mailbox even if the device itself had
gone away.
To address the responsiveness issues we can decrease the sleeps to 20ms and
use a jiffies based timeout value rather than just counting the number of
times we slept and then polled.
To address the hardware going away we can move the check for the firmware
BAR being present from where it was and place it inside the loop after the
mailbox descriptor ring is initialized and before we sleep so that we just
abort and return an error if the device went away during initialization.
With these two changes we see a significant improvement in boot times for
the driver.
Fixes: da3cde08209e ("eth: fbnic: Add FW communication mechanism")
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/174654721224.499179.2698616208976624755.stgit@ahduyck-xeon-server.home.arpa
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
There was an issue in that if we were to shutdown we could be left with
a completion in flight as the mailbox went away. To address that I have
added an fbnic_mbx_evict_all_cmpl function that is meant to essentially
create a "broken pipe" type response so that all callers will receive an
error indicating that the connection has been broken as a result of us
shutting down the mailbox.
Fixes: 378e5cc1c6c6 ("eth: fbnic: hwmon: Add completion infrastructure for firmware requests")
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/174654720578.499179.380252598204530873.stgit@ahduyck-xeon-server.home.arpa
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
The fbnic_mbx_flush_tx function had a number of issues.
First, we were waiting 200ms for the firmware to process the packets. We
can drop this to 20ms and in almost all cases this should be more than
enough time. So by changing this we can significantly reduce shutdown time.
Second, we were not making sure that the Tx path was actually shut off. As
such we could still have packets added while we were flushing the mailbox.
To prevent that we can now clear the ready flag for the Tx side and it
should stay down since the interrupt is disabled.
Third, we kept re-reading the tail due to the second issue. The tail should
not move after we have started the flush so we can just read it once while
we are holding the mailbox Tx lock. By doing that we are guaranteed that
the value should be consistent.
Fourth, we were keeping a count of descriptors cleaned due to the second
and third issues called out. That count is not a valid reason to be exiting
the cleanup, and with the tail only being read once we shouldn't see any
cases where the tail moves after the disable so the tracking of count can
be dropped.
Fifth, we were using attempts * sleep time to determine how long we would
wait in our polling loop to flush out the Tx. This can be very imprecise.
In order to tighten up the timing we are shifting over to using a jiffies
value of jiffies + 10 * HZ + 1 to determine the jiffies value we should
stop polling at as this should be accurate within once sleep cycle for the
total amount of time spent polling.
Fixes: da3cde08209e ("eth: fbnic: Add FW communication mechanism")
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/174654719929.499179.16406653096197423749.stgit@ahduyck-xeon-server.home.arpa
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
We have two issues that need to be addressed in our IRQ handling.
One is the fact that we can end up double-freeing IRQs in the event of an
exception handling error such as a PCIe reset/recovery that fails. To
prevent that from becoming an issue we can use the msix_vector values to
indicate that we have successfully requested/freed the IRQ by only setting
or clearing them when we have completed the given action.
The other issue is that we have several potential races in our IRQ path due
to us manipulating the mask before the vector has been truly disabled. In
order to handle that in the case of the FW mailbox we need to not
auto-enable the IRQ and instead will be enabling/disabling it separately.
In the case of the PCS vector we can mitigate this by unmapping it and
synchronizing the IRQ before we clear the mask.
The general order of operations after this change is now to request the
interrupt, poll the FW mailbox to ready, and then enable the interrupt. For
the shutdown we do the reverse where we disable the interrupt, flush any
pending Tx, and then free the IRQ. I am renaming the enable/disable to
request/free to be equivilent with the IRQ calls being used. We may see
additions in the future to enable/disable the IRQs versus request/free them
for certain use cases.
Fixes: da3cde08209e ("eth: fbnic: Add FW communication mechanism")
Fixes: 69684376eed5 ("eth: fbnic: Add link detection")
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/174654719271.499179.3634535105127848325.stgit@ahduyck-xeon-server.home.arpa
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
In order to prevent the device from throwing spurious writes and/or reads
at us we need to gate the AXI fabric interface to the PCIe until such time
as we know the FW is in a known good state.
To accomplish this we use the mailbox as a mechanism for us to recognize
that the FW has acknowledged our presence and is no longer sending any
stale message data to us.
We start in fbnic_mbx_init by calling fbnic_mbx_reset_desc_ring function,
disabling the DMA in both directions, and then invalidating all the
descriptors in each ring.
We then poll the mailbox in fbnic_mbx_poll_tx_ready and when the interrupt
is set by the FW we pick it up and mark the mailboxes as ready, while also
enabling the DMA.
Once we have completed all the transactions and need to shut down we call
into fbnic_mbx_clean which will in turn call fbnic_mbx_reset_desc_ring for
each ring and shut down the DMA and once again invalidate the descriptors.
Fixes: 3646153161f1 ("eth: fbnic: Add register init to set PCIe/Ethernet device config")
Fixes: da3cde08209e ("eth: fbnic: Add FW communication mechanism")
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/174654718623.499179.7445197308109347982.stgit@ahduyck-xeon-server.home.arpa
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Address to issues with the FW mailbox descriptor initialization.
We need to reverse the order of accesses when we invalidate an entry versus
writing an entry. When writing an entry we write upper and then lower as
the lower 32b contain the valid bit that makes the entire address valid.
However for invalidation we should write it in the reverse order so that
the upper is marked invalid before we update it.
Without this change we may see FW attempt to access pages with the upper
32b of the address set to 0 which will likely result in DMAR faults due to
write access failures on mailbox shutdown.
Fixes: da3cde08209e ("eth: fbnic: Add FW communication mechanism")
Signed-off-by: Alexander Duyck <alexanderduyck@fb.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/174654717972.499179.8083789731819297034.stgit@ahduyck-xeon-server.home.arpa
Reviewed-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
|
|
Jonas Gorski says:
====================
net: dsa: b53: accumulated fixes
This patchset aims at fixing most issues observed while running the
vlan_unaware_bridge, vlan_aware_bridge and local_termination selftests.
Most tests succeed with these patches on BCM53115, connected to a
BCM6368.
It took me a while to figure out that a lot of tests will fail if all
ports have the same MAC address, as the switches drop any frames with
DA == SA. Luckily BCM63XX boards often have enough MACs allocated for
all ports, so I just needed to assign them.
The still failing tests are:
FDB learning, both vlan aware aware and unaware:
This is expected, as b53 currently does not implement changing the
ageing time, and both the bridge code and DSA ignore that, so the
learned entries don't age out as expected.
ping and ping6 in vlan unaware:
These fail because of the now fixed learning, the switch trying to
forward packet ingressing on one of the standalone ports to the learned
port of the mac address when the packets ingressed on the bridged port.
The port VLAN masks only prevent forwarding to other ports, but the ARL
lookup will still happen, and the packet gets dropped because the port
isn't allowed to forward there.
I have a fix/workaround for that, but as it is a bit more controversial
and makes use of an unrelated feature, I decided to hold off from that
and post it later.
This wasn't noticed so far, because learning was never working in VLAN
unaware mode, so the traffic was always broadcast (which sidesteps the
issue).
Finally some of the multicast tests from local_termination fail, where
the reception worked except it shouldn't. This doesn't seem to me as a
super serious issue, so I didn't attempt to debug/fix these yet.
I'm not super confident I didn't break sf2 along the way, but I did
compile test and tried to find ways it cause issues (I failed to find
any). I hope Florian will tell me.
====================
Link: https://patch.msgid.link/20250429201710.330937-1-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When a port gets set up, b53 disables learning and enables the port for
flooding. This can undo any bridge configuration on the port.
E.g. the following flow would disable learning on a port:
$ ip link add br0 type bridge
$ ip link set sw1p1 master br0 <- enables learning for sw1p1
$ ip link set br0 up
$ ip link set sw1p1 up <- disables learning again
Fix this by populating dsa_switch_ops::port_setup(), and set up initial
config there.
Fixes: f9b3827ee66c ("net: dsa: b53: Support setting learning on port")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-12-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When VLAN filtering is off, we configure the switch to forward, but not
learn on VLAN table misses. This effectively disables learning while not
filtering.
Fix this by switching to forward and learn. Setting the learning disable
register will still control whether learning actually happens.
Fixes: dad8d7c6452b ("net: dsa: b53: Properly account for VLAN filtering")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-11-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
To allow runtime switching between vlan aware and vlan non-aware mode,
we need to properly keep track of any bridge VLAN configuration.
Likewise, we need to know when we actually switch between both modes, to
not have to rewrite the full VLAN table every time we update the VLANs.
So keep track of the current vlan_filtering mode, and on changes, apply
the appropriate VLAN configuration.
Fixes: 0ee2af4ebbe3 ("net: dsa: set configure_vlan_while_not_filtering to true by default")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-10-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Documentation/networking/switchdev.rst says:
- with VLAN filtering turned off: the bridge is strictly VLAN unaware and its
data path will process all Ethernet frames as if they are VLAN-untagged.
The bridge VLAN database can still be modified, but the modifications should
have no effect while VLAN filtering is turned off.
This breaks if we immediately apply the VLAN configuration, so skip
writing it when vlan_filtering is off.
Fixes: 0ee2af4ebbe3 ("net: dsa: set configure_vlan_while_not_filtering to true by default")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-9-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Since we cannot set forwarding destinations per VLAN, we should not have
a VLAN 0 configured, as it would allow untagged traffic to work across
ports on VLAN aware bridges regardless if a PVID untagged VLAN exists.
So remove the VLAN 0 on join, an re-add it on leave. But only do so if
we have a VLAN aware bridge, as without it, untagged traffic would
become tagged with VID 0 on a VLAN unaware bridge.
Fixes: a2482d2ce349 ("net: dsa: b53: Plug in VLAN support")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-8-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
While JOIN_ALL_VLAN allows to join all VLANs, we still need to keep the
default VLAN enabled so that untagged traffic stays untagged.
So rejoin the default VLAN even for switches with JOIN_ALL_VLAN support.
Fixes: 48aea33a77ab ("net: dsa: b53: Add JOIN_ALL_VLAN support")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-7-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The untagged default VLAN is added to the default vlan, which may be
one, but we modify the VLAN 0 entry on bridge leave.
Fix this to use the correct VLAN entry for the default pvid.
Fixes: fea83353177a ("net: dsa: b53: Fix default VLAN ID")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-6-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Presumably the intention here was to flush the VLAN of the old pvid, not
the added VLAN again, which we already flushed before.
Fixes: a2482d2ce349 ("net: dsa: b53: Plug in VLAN support")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-5-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Currently the PVID of ports are only set when adding/updating VLANs with
PVID set or removing VLANs, but not when clearing the PVID flag of a
VLAN.
E.g. the following flow
$ ip link add br0 type bridge vlan_filtering 1
$ ip link set sw1p1 master bridge
$ bridge vlan add dev sw1p1 vid 10 pvid untagged
$ bridge vlan add dev sw1p1 vid 10 untagged
Would keep the PVID set as 10, despite the flag being cleared. Fix this
by checking if we need to unset the PVID on vlan updates.
Fixes: a2482d2ce349 ("net: dsa: b53: Plug in VLAN support")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-4-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
The Broadcom management header does not carry the original VLAN tag
state information, just the ingress port, so for untagged frames we do
not know from which VLAN they originated.
Therefore keep the CPU port always tagged except for VLAN 0.
Fixes the following setup:
$ ip link add br0 type bridge vlan_filtering 1
$ ip link set sw1p1 master br0
$ bridge vlan add dev br0 pvid untagged self
$ ip link add sw1p2.10 link sw1p2 type vlan id 10
Where VID 10 would stay untagged on the CPU port.
Fixes: 2c32a3d3c233 ("net: dsa: b53: Do not force CPU to be always tagged")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-3-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Allow reserved multicast to ignore VLAN membership so STP and other
management protocols work without a PVID VLAN configured when using a
vlan aware bridge.
Fixes: 967dd82ffc52 ("net: dsa: b53: Add support for Broadcom RoboSwitch")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Tested-by: Florian Fainelli <florian.fainelli@broadcom.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Link: https://patch.msgid.link/20250429201710.330937-2-jonas.gorski@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Meghana Malladi says:
====================
Bug fixes from XDP patch series
This patch series fixes the bugs introduced while adding
xdp support in the icssg driver, and were reproduced while
running xdp-trafficgen to generate xdp traffic on icssg interfaces.
v1: https://lore.kernel.org/all/20250428120459.244525-1-m-malladi@ti.com/
====================
Link: https://patch.msgid.link/20250506110546.4065715-1-m-malladi@ti.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When sending out any kind of traffic, it is essential that the driver
keeps reporting BQL of the number of bytes that have been sent so that
BQL can track the amount of data in the queue and prevents it from
overflowing. If BQL is not reported, the driver may continue sending
packets even when the queue is full, leading to packet loss, congestion
and decreased network performance. Currently this is missing in
emac_xmit_xdp_frame() and this patch fixes it.
Fixes: 62aa3246f462 ("net: ti: icssg-prueth: Add XDP support")
Signed-off-by: Meghana Malladi <m-malladi@ti.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20250506110546.4065715-4-m-malladi@ti.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
Add __netif_tx_lock() to ensure that only one packet is being
transmitted at a time to avoid race conditions in the netif_txq
struct and prevent packet data corruption. Failing to do so causes
kernel panic with the following error:
[ 2184.746764] ------------[ cut here ]------------
[ 2184.751412] kernel BUG at lib/dynamic_queue_limits.c:99!
[ 2184.756728] Internal error: Oops - BUG: 00000000f2000800 [#1] PREEMPT SMP
logs: https://gist.github.com/MeghanaMalladiTI/9c7aa5fc3b7fb03f87c74aad487956e9
The lock is acquired before calling emac_xmit_xdp_frame() and released after the
call returns. This ensures that the TX queue is protected from concurrent access
during the transmission of XDP frames.
Fixes: 62aa3246f462 ("net: ti: icssg-prueth: Add XDP support")
Signed-off-by: Meghana Malladi <m-malladi@ti.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20250506110546.4065715-3-m-malladi@ti.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
xdp_features demonstrates what all XDP capabilities are supported
on a given network device. The driver needs to set these xdp_features
flag to let the network stack know what XDP features a given driver
is supporting. These flags need to be set for a given ndev irrespective
of any XDP program being loaded or not.
Fixes: 62aa3246f462 ("net: ti: icssg-prueth: Add XDP support")
Signed-off-by: Meghana Malladi <m-malladi@ti.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Link: https://patch.msgid.link/20250506110546.4065715-2-m-malladi@ti.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When switching network namespaces with the bpf_redirect_peer helper, the
skb->mark and skb->tstamp fields are not zeroed out like they can be on
a typical netns switch. This patch clarifies that in the helper
description.
Signed-off-by: Paul Chaignon <paul.chaignon@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://patch.msgid.link/ccc86af26d43c5c0b776bcba2601b7479c0d46d0.1746460653.git.paul.chaignon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
|
When bpf_redirect_peer is used to redirect packets to a device in
another network namespace, the skb isn't scrubbed. That can lead skb
information from one namespace to be "misused" in another namespace.
As one example, this is causing Cilium to drop traffic when using
bpf_redirect_peer to redirect packets that just went through IPsec
decryption to a container namespace. The following pwru trace shows (1)
the packet path from the host's XFRM layer to the container's XFRM
layer where it's dropped and (2) the number of active skb extensions at
each function.
NETNS MARK IFACE TUPLE FUNC
4026533547 d00 eth0 10.244.3.124:35473->10.244.2.158:53 xfrm_rcv_cb
.active_extensions = (__u8)2,
4026533547 d00 eth0 10.244.3.124:35473->10.244.2.158:53 xfrm4_rcv_cb
.active_extensions = (__u8)2,
4026533547 d00 eth0 10.244.3.124:35473->10.244.2.158:53 gro_cells_receive
.active_extensions = (__u8)2,
[...]
4026533547 0 eth0 10.244.3.124:35473->10.244.2.158:53 skb_do_redirect
.active_extensions = (__u8)2,
4026534999 0 eth0 10.244.3.124:35473->10.244.2.158:53 ip_rcv
.active_extensions = (__u8)2,
4026534999 0 eth0 10.244.3.124:35473->10.244.2.158:53 ip_rcv_core
.active_extensions = (__u8)2,
[...]
4026534999 0 eth0 10.244.3.124:35473->10.244.2.158:53 udp_queue_rcv_one_skb
.active_extensions = (__u8)2,
4026534999 0 eth0 10.244.3.124:35473->10.244.2.158:53 __xfrm_policy_check
.active_extensions = (__u8)2,
4026534999 0 eth0 10.244.3.124:35473->10.244.2.158:53 __xfrm_decode_session
.active_extensions = (__u8)2,
4026534999 0 eth0 10.244.3.124:35473->10.244.2.158:53 security_xfrm_decode_session
.active_extensions = (__u8)2,
4026534999 0 eth0 10.244.3.124:35473->10.244.2.158:53 kfree_skb_reason(SKB_DROP_REASON_XFRM_POLICY)
.active_extensions = (__u8)2,
In this case, there are no XFRM policies in the container's network
namespace so the drop is unexpected. When we decrypt the IPsec packet,
the XFRM state used for decryption is set in the skb extensions. This
information is preserved across the netns switch. When we reach the
XFRM policy check in the container's netns, __xfrm_policy_check drops
the packet with LINUX_MIB_XFRMINNOPOLS because a (container-side) XFRM
policy can't be found that matches the (host-side) XFRM state used for
decryption.
This patch fixes this by scrubbing the packet when using
bpf_redirect_peer, as is done on typical netns switches via veth
devices except skb->mark and skb->tstamp are not zeroed.
Fixes: 9aa1206e8f482 ("bpf: Add redirect_peer helper")
Signed-off-by: Paul Chaignon <paul.chaignon@gmail.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Martin KaFai Lau <martin.lau@kernel.org>
Link: https://patch.msgid.link/1728ead5e0fe45e7a6542c36bd4e3ca07a73b7d6.1746460653.git.paul.chaignon@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
|
