netdev - [PATCH iwl-net v2 0/5] idpf: fix issues in the reset handling path

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-Id: <20251121001218.4565-1-emil.s.tantilov@intel.com>
Date: Thu, 20 Nov 2025 16:12:13 -0800
From: Emil Tantilov <emil.s.tantilov@...el.com>
To: intel-wired-lan@...ts.osuosl.org
Cc: netdev@...r.kernel.org,
	Aleksandr.Loktionov@...el.com,
	przemyslaw.kitszel@...el.com,
	anthony.l.nguyen@...el.com,
	andrew+netdev@...n.ch,
	davem@...emloft.net,
	edumazet@...gle.com,
	kuba@...nel.org,
	pabeni@...hat.com,
	decot@...gle.com,
	willemb@...gle.com,
	joshua.a.hay@...el.com,
	madhu.chittim@...el.com,
	aleksander.lobakin@...el.com,
	larysa.zaremba@...el.com,
	iamvivekkumar@...gle.com
Subject: [PATCH iwl-net v2 0/5] idpf: fix issues in the reset handling path

Resolve multiple issues in the error path during reset handling of the
IDPF driver:
- Mailbox timeouts in the init task during a reset lead to the netdevs
  being unregistered.
- Reset times out and leaves the netdevs exposed to callbacks with the
  vport resources already freed.
- Simultaneous reset and soft reset calls will result in the loss of the
  vport state, leaving the netdev in DOWN state.
- Memory leak in idpf_vport_rel() where vport->rx_ptype_lkup was not
  freed during a reset.
- Memory leak in idpf_vc_core_deinit(), where kfree() was missing for
  hw->lan_regs.
- Crash on reset following a failed init on load, due to the service
  tasks still running in the background and attempting to handle the
  reset while resources are not initialized.

Changelog:
v2:
 - Patch 1 - No change.
 - Patch 2 - Changed the check for vport being NULL by accessing it via the
   adapter struct instead of idpf_netdev_priv. This avoids potential NULL
   pointer in the case where the init task failed on driver load.
   Updated comments to clarify vports can be NULL before the call to
   idpf_attach_and_open().
 - Patch 3 - No change.
 - Patch 4 - No change.
 - Patch 5 - New to the series, to fix a crash, following a failed driver
   load due to an error in the init task.

v1: https://lore.kernel.org/intel-wired-lan/20251110180823.18008-1-emil.s.tantilov@intel.com/

Emil Tantilov (5):
  idpf: keep the netdev when a reset fails
  idpf: detach and close netdevs while handling a reset
  idpf: fix memory leak in idpf_vport_rel()
  idpf: fix memory leak in idpf_vc_core_deinit()
  idpf: fix error handling in the init_task on load

 drivers/net/ethernet/intel/idpf/idpf_lib.c    | 156 +++++++++++-------
 .../net/ethernet/intel/idpf/idpf_virtchnl.c   |   4 +
 2 files changed, 96 insertions(+), 64 deletions(-)

-- 
2.37.3