linux-kernel - Re: 3.8-rc2/rc3 write() blocked on CLOSE

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]

Date:	Thu, 10 Jan 2013 18:01:15 -0800
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	Eric Wong <normalperson@...t.net>
Cc:	netdev@...r.kernel.org, Mel Gorman <mgorman@...e.de>,
	linux-mm@...ck.org, linux-kernel@...r.kernel.org,
	Rik van Riel <riel@...hat.com>,
	Minchan Kim <minchan@...nel.org>,
	Andrew Morton <akpm@...ux-foundation.org>,
	Linus Torvalds <torvalds@...ux-foundation.org>
Subject: Re: 3.8-rc2/rc3 write() blocked on CLOSE_WAIT TCP socket

On Fri, 2013-01-11 at 00:49 +0000, Eric Wong wrote:
> The below Ruby script reproduces the issue for me with write()
> getting stuck, usually with a few iterations (sometimes up to 100).
> 
> I've reproduced this with 3.8-rc2 and rc3, even with Mel's partial
> revert patch in <20130110194212.GJ13304@...e.de> applied.
> 
> I can not reproduce this with 3.7.1+
>    stable-queue 2afd72f59c518da18853192ceeebead670ced5ea
> So this seems to be a new bug from the 3.8 cycle...
> 
> Fortunately, this bug far easier for me to reproduce than the ppoll+send
> (toosleepy) failures.
> 
> Both socat and ruby (Ruby 1.8, 1.9, 2.0 should all work), along with
> common shell tools (dd, sh, cat) are required for testing this:
> 
> 	# 100 iterations, raise/lower the number if needed
> 	ruby the_script_below.rb 100
> 
> lsof -p 15236 reveals this:
> ruby    15236   ew    5u  IPv4  23066      0t0     TCP localhost:33728->localhost:38658 (CLOSE_WAIT)

Hmm, it might be commit c3ae62af8e755ea68380fb5ce682e60079a4c388
tcp: should drop incoming frames without ACK flag set

It seems RST should be allowed to not have ACK set.

I'll send a fix, thanks !





--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/