Recent versions of the E810 firmware have support for a "low latency"
interface for accessing Tx timestamps without interacting over the AdminQ.
Unfortunately, a couple of the functions involved with this logic don't
properly check if the Tx timestamp tracker is initialized. This can lead to
a use-after-free or NULL pointer access violation:
[245977.278756] BUG: kernel NULL pointer dereference, address: 0000000000000000
[245977.278774] RIP: 0010:_find_first_bit+0x19/0x40
[245977.278796] Call Trace:
[245977.278809] ? ice_misc_intr+0x364/0x380 [ice]
This happens because the reset flow in the driver re-initializes the
tracker. If a Tx timestamp interrupt occurs concurrently with a reset, the
ineractions race and the crash can occur.
I split the fixes to two separate patches because the extra interrupt for
low latency came in a different kernel. Hopefully this aids with
backporting the fix.
This was originally reported on an older kernel prior to f9472aaabd1f
("ice: Process TSYN IRQ in a separate function"), which has a similar bug,
but at a different code point. I plan to look into if any stable kernels
need a separate fix after this.
Signed-off-by: Jacob Keller <[email protected]>
---
Jacob Keller (2):
ice: fix NULL access of tx->in_use in ice_ptp_ts_irq
ice: fix NULL access of tx->in_use in ice_ll_ts_intr
drivers/net/ethernet/intel/ice/ice_main.c | 12 +++++++-----
drivers/net/ethernet/intel/ice/ice_ptp.c | 13 ++++++++-----
2 files changed, 15 insertions(+), 10 deletions(-)
---
base-commit: d942fe13f72bec92f6c689fbd74c5ec38228c16a
change-id: 20250807-jk-ice-fix-tx-tstamp-race-5fe8484670a4
Best regards,
--
Jacob Keller <[email protected]>
