Received-SPF: pass (google.com: domain of linux-kernel+bounces-109540-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Date: Thu, 21 Mar 2024 10:36:21 +0900 (JST)
Message-Id: <20240321.103621.1433494402179493599.syoshida@redhat.com>
To: nicolas.cavallari@green-communications.fr
Cc: davem@davemloft.net, dsahern@kernel.org, edumazet@google.com,
 kuba@kernel.org, pabeni@redhat.com, netdev@vger.kernel.org,
 linux-kernel@vger.kernel.org
Subject: Re: [PATCH] ipv6: delay procfs initialization after the ipv6
 structs are ready
From: Shigeru Yoshida <syoshida@redhat.com>
In-Reply-To: <20240320171858.2671-1-nicolas.cavallari@green-communications.fr>
References: <20240320171858.2671-1-nicolas.cavallari@green-communications.fr>
Precedence: bulk
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

On Wed, 20 Mar 2024 18:17:36 +0100, Nicolas Cavallari wrote:
> procfs files are created before the structure they reference are
> initialized.  For example, if6_proc_init() creates procfs files that
> access structures initialized by addrconf_init().
> 
> If ipv6 is compiled as a module and a program manages to open an ipv6
> procfs file during the loading of the module, it can oops the kernel.
> 
> It appears that we were unlucky enough to reproduce this problem
> multiple times already, out of maybe 100 boots:
> 
> NET: Registered PF_INET6 protocol family
> 8<--- cut here ---
> pwm-backlight backlight: supply power not found, using dummy regulator
> Segment Routing with IPv6
> In-situ OAM (IOAM) with IPv6
> Unable to handle kernel NULL pointer dereference at virtual address
>  00000000
> mt7915e 0000:03:00.0 wlp3s0: renamed from wlan0
> [00000000] *pgd=00000000
> Internal error: Oops: 5 [#1] SMP ARM
> Modules linked in: ipv6 mt7915e mt76_connac_lib mt76 dw_hdmi_imx
>  mac80211 dw_hdmi drm_display_helper imxdrm drm_dma_helper
>  drm_kms_helper snd_soc_imx_sgtl5000 syscopyarea sysfillrect sysimgblt
>  fb_sys_fops imx_ipu_v3 snd_soc_fsl_asoc_card cfg80211 snd_soc_sgtl5000
>  drm libarc4 snd_soc_fsl_ssi snd_soc_simple_card_utils imx_pcm_dma
>  snd_soc_core rfkill snd_pcm_dmaengine snd_pcm
>  drm_panel_orientation_quirks cfbfillrect cfbimgblt cfbcopyarea
>  snd_timer snd egalax_ts snd_soc_imx_audmux soundcore flexcan mux_mmio
>  imx2_wdt mux_core can_dev pwm_bl
> CPU: 2 PID: 850 Comm: snmpd Not tainted 6.1.14 #1
> Hardware name: Freescale i.MX6 Quad/DualLite (Device Tree)
> PC is at if6_seq_start+0x2c/0x98 [ipv6]
> LR is at init_net+0x0/0xc00
> [...]
>  if6_seq_start [ipv6] from seq_read_iter+0xb4/0x510
>  seq_read_iter from seq_read+0x80/0xac
>  seq_read from proc_reg_read+0xac/0x100
>  proc_reg_read from vfs_read+0xb0/0x284
>  vfs_read from ksys_read+0x64/0xec
>  ksys_read from ret_fast_syscall+0x0/0x54
> Exception stack(0xf0e31fa8 to 0xf0e31ff0)
> 1fa0:                   b67fd0b0 be8a666b 0000000a b67fd148 00000400
>  00000000
> 1fc0: b67fd0b0 be8a666b 00000001 00000003 be8a67ec 00000000 b6d7e000
>  b6c9954a
> 1fe0: b6d7eb30 be8a6638 b6ef11b4 b6ef0ddc
> Code: e5931004 e35100ff ca000014 e59e25dc (e7920101)
> ---[ end trace 0000000000000000 ]---
> 
> Signed-off-by: Nicolas Cavallari <nicolas.cavallari@green-communications.fr>

The following tag is needed?

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")

Also, we should put appropriate prefix in the subject for netdev
patches, i.e. "net" or "net-next". As for this patch, I think
appropriate prefix is "[PATCH net]".

Anyway,

Reviewed-by: Shigeru Yoshida <syoshida@redhat.com>

Thanks,
Shigeru