netdev - Re: [Intel-wired-lan] [PATCH iwl-net v2] ice: fix memleak in ice_init_tx

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 for Android: free password hash cracker in your pocket

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

Message-ID: <4fc61caf-e922-44d6-b3b6-f286fe179107@intel.com>
Date: Tue, 10 Sep 2024 14:30:34 -0700
From: Jacob Keller <jacob.e.keller@...el.com>
To: Przemek Kitszel <przemyslaw.kitszel@...el.com>,
	<intel-wired-lan@...ts.osuosl.org>, Tony Nguyen <anthony.l.nguyen@...el.com>
CC: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@...el.com>, Larysa Zaremba
	<larysa.zaremba@...el.com>, <netdev@...r.kernel.org>, Mateusz Polchlopek
	<mateusz.polchlopek@...el.com>
Subject: Re: [Intel-wired-lan] [PATCH iwl-net v2] ice: fix memleak in
 ice_init_tx_topology()



On 9/10/2024 6:57 AM, Przemek Kitszel wrote:
> Fix leak of the FW blob (DDP pkg).
> 
> Make ice_cfg_tx_topo() const-correct, so ice_init_tx_topology() can avoid
> copying whole FW blob. Copy just the topology section, and only when
> needed. Reuse the buffer allocated for the read of the current topology.
> 
> This was found by kmemleak, with the following trace for each PF:
>     [<ffffffff8761044d>] kmemdup_noprof+0x1d/0x50
>     [<ffffffffc0a0a480>] ice_init_ddp_config+0x100/0x220 [ice]
>     [<ffffffffc0a0da7f>] ice_init_dev+0x6f/0x200 [ice]
>     [<ffffffffc0a0dc49>] ice_init+0x29/0x560 [ice]
>     [<ffffffffc0a10c1d>] ice_probe+0x21d/0x310 [ice]
> 
> Constify ice_cfg_tx_topo() @buf parameter.
> This cascades further down to few more functions.
> 

Nice!

Reviewed-by: Jacob Keller <jacob.e.keller@...el.com>

> Fixes: cc5776fe1832 ("ice: Enable switching default Tx scheduler topology")
> CC: Larysa Zaremba <larysa.zaremba@...el.com>
> CC: Jacob Keller <jacob.e.keller@...el.com>
> CC: Pucha Himasekhar Reddy <himasekharx.reddy.pucha@...el.com>
> CC: Mateusz Polchlopek <mateusz.polchlopek@...el.com>
> Signed-off-by: Przemek Kitszel <przemyslaw.kitszel@...el.com>
> ---
> this patch obsoletes two other, so I'm dropping RB tags
> v1, iwl-net: https://lore.kernel.org/netdev/20240904123306.14629-2-przemyslaw.kitszel@intel.com/
>     wrong assumption that ice_get_set_tx_topo() does not modify the buffer
>     on adminq SET variant, it sometimes does, to fill the response, thanks
>     to Pucha Himasekhar Reddy for finding it out;
> almost-const-correct iwl-next patch:
> https://lore.kernel.org/intel-wired-lan/20240904093135.8795-2-przemyslaw.kitszel@intel.com
> it was just some of this patch, now it is const-correct
> ---

Right. So now we're doing the const correctness in this patch along with
the fix?

Would it make sense to fix the copy issue but leave const updates to the
next tree?

I think I'm fine with this, but wonder if it will make backporting a bit
more difficult? Probably not, given that this code is rarely modified.

The const fixes are also relatively smaller than I anticipated :D

Thanks,
Jake