]> www.infradead.org Git - users/jedix/linux-maple.git/commit
ice: fix Tx scheduler error handling in XDP callback
authorMichal Kubiak <michal.kubiak@intel.com>
Tue, 13 May 2025 10:55:27 +0000 (12:55 +0200)
committerTony Nguyen <anthony.l.nguyen@intel.com>
Fri, 30 May 2025 20:54:17 +0000 (13:54 -0700)
commit0153f36041b8e52019ebfa8629c13bf8f9b0a951
tree14eba39b0504df269e6b426e22621a0ef380200e
parentd3faab9b5a6a0477d69c38bd11c43aa5e936f929
ice: fix Tx scheduler error handling in XDP callback

When the XDP program is loaded, the XDP callback adds new Tx queues.
This means that the callback must update the Tx scheduler with the new
queue number. In the event of a Tx scheduler failure, the XDP callback
should also fail and roll back any changes previously made for XDP
preparation.

The previous implementation had a bug that not all changes made by the
XDP callback were rolled back. This caused the crash with the following
call trace:

[  +9.549584] ice 0000:ca:00.0: Failed VSI LAN queue config for XDP, error: -5
[  +0.382335] Oops: general protection fault, probably for non-canonical address 0x50a2250a90495525: 0000 [#1] SMP NOPTI
[  +0.010710] CPU: 103 UID: 0 PID: 0 Comm: swapper/103 Not tainted 6.14.0-net-next-mar-31+ #14 PREEMPT(voluntary)
[  +0.010175] Hardware name: Intel Corporation M50CYP2SBSTD/M50CYP2SBSTD, BIOS SE5C620.86B.01.01.0005.2202160810 02/16/2022
[  +0.010946] RIP: 0010:__ice_update_sample+0x39/0xe0 [ice]

[...]

[  +0.002715] Call Trace:
[  +0.002452]  <IRQ>
[  +0.002021]  ? __die_body.cold+0x19/0x29
[  +0.003922]  ? die_addr+0x3c/0x60
[  +0.003319]  ? exc_general_protection+0x17c/0x400
[  +0.004707]  ? asm_exc_general_protection+0x26/0x30
[  +0.004879]  ? __ice_update_sample+0x39/0xe0 [ice]
[  +0.004835]  ice_napi_poll+0x665/0x680 [ice]
[  +0.004320]  __napi_poll+0x28/0x190
[  +0.003500]  net_rx_action+0x198/0x360
[  +0.003752]  ? update_rq_clock+0x39/0x220
[  +0.004013]  handle_softirqs+0xf1/0x340
[  +0.003840]  ? sched_clock_cpu+0xf/0x1f0
[  +0.003925]  __irq_exit_rcu+0xc2/0xe0
[  +0.003665]  common_interrupt+0x85/0xa0
[  +0.003839]  </IRQ>
[  +0.002098]  <TASK>
[  +0.002106]  asm_common_interrupt+0x26/0x40
[  +0.004184] RIP: 0010:cpuidle_enter_state+0xd3/0x690

Fix this by performing the missing unmapping of XDP queues from
q_vectors and setting the XDP rings pointer back to NULL after all those
queues are released.
Also, add an immediate exit from the XDP callback in case of ring
preparation failure.

Fixes: efc2214b6047 ("ice: Add support for XDP")
Reviewed-by: Dawid Osuchowski <dawid.osuchowski@linux.intel.com>
Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
Reviewed-by: Jacob Keller <jacob.e.keller@intel.com>
Signed-off-by: Michal Kubiak <michal.kubiak@intel.com>
Reviewed-by: Aleksandr Loktionov <aleksandr.loktionov@intel.com>
Reviewed-by: Simon Horman <horms@kernel.org>
Tested-by: Jesse Brandeburg <jbrandeburg@cloudflare.com>
Tested-by: Saritha Sanigani <sarithax.sanigani@intel.com> (A Contingent Worker at Intel)
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
drivers/net/ethernet/intel/ice/ice_main.c