netdev - Re: [PATCH] bnx2x: avoid atomic allocations during initialization

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

Message-ID: <CAM8tLiPCUS3_jo4KU1zY9+8wRg08fA0ohH9d+0a4jyC69D9JZg@mail.gmail.com>
Date:	Sat, 7 Sep 2013 12:07:50 +0300
From:	Dmitry Kravkov <dkravkov@...il.com>
To:	David Miller <davem@...emloft.net>
Cc:	Dmitry Kravkov <dmitry@...adcom.com>,
	Tereza Schmidtová <mschmidt@...hat.com>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>,
	Ariel Elior <ariele@...adcom.com>,
	Eilon Greenstein <eilong@...adcom.com>
Subject: Re: [PATCH] bnx2x: avoid atomic allocations during initialization

On Sat, Sep 7, 2013 at 7:29 AM, David Miller <davem@...emloft.net> wrote:
> From: "Dmitry Kravkov" <dmitry@...adcom.com>
> Date: Sat, 7 Sep 2013 01:45:10 +0000
>
>> Once you allocated the memory during initialization , you will most
>> probably fail to allocate its replacement during RX handling (on
>> this machine).
>
> Not true, these two events can exist in totally different timeframes
> and totally different memory pressure situations.
As seen from the customers, these allocation fail when system
constantly under memory pressure - like VM environment or 32-bit
platforms

>
> Probe should never fail because of an atomic memory allocation if at
> all possible.
We here are in open() flow, not probe()

>
> You should use sleeping allocations in every possible situation for
> maximum stability, and this absolutely matches the approach taken
> by every other major driver.
Sleeping allocation will be valid only for one ring overlap - it just
delays the issue, not resolving it, after ring overlap ALL of them
will be replaced by allocations from interrupt context. And again if
we, discover issue earlier we can provide system with fully working
networking with very acceptable performance using SW GRO stack.
Instead of having device which drops every TPA/FW_GRO aggregation -
user will not able to ssh :(

> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@...r.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html