netdev - "swiotlb buffer is full" with 3.13-rc1+ but not 3.4.

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]

Date:	Fri, 29 Nov 2013 16:04:17 +0000
From:	Ian Jackson <Ian.Jackson@...citrix.com>
To:	<netdev@...r.kernel.org>, Michael Chan <mchan@...adcom.com>,
	<dl-mptfusionlinux@....com>, <linux-scsi@...r.kernel.org>,
	<support@....com>, Sreekanth Reddy <Sreekanth.Reddy@....com>,
	Nagalakshmi Nandigama <Nagalakshmi.Nandigama@....com>,
	Konrad Rzeszutek Wilk <konrad.wilk@...cle.com>,
	<xen-devel@...ts.xenproject.org>, <linux-kernel@...r.kernel.org>
Subject: "swiotlb buffer is full" with 3.13-rc1+ but not 3.4.

I have a pair of machines (AMD) which, under Xen, produces a lot of
these messages and stops working:

 [ 1892.817406] mptsas 0000:03:00.0: swiotlb buffer is full
 [ 1892.817425] mptsas 0000:03:00.0: swiotlb buffer is full
 [ 1892.835403] mptsas 0000:03:00.0: swiotlb buffer is full

and in another run

 [  448.434824] bnx2 0000:02:00.0: swiotlb buffer is full
 [  448.676476] bnx2 0000:02:00.0: swiotlb buffer is full
 [  448.677237] bnx2 0000:02:00.0: swiotlb buffer is full

This is with e7c0d3dac88b ("v3.13-rc1-229-ge7c0d3d").  3.4.x works
fine.  I'm not sure exactly when it started happening; it seems to
only happen after some work is done on the host.  The bug is
host-specific, but appears to happen on both of my two identical
machines of this configuration, so probably isn't a hardware fault.
The hypervisor version is xen.git#e439e0b289e3.

The logs from a recent failure are here:
  http://www.chiark.greenend.org.uk/~xensrcts/logs/22171/test-amd64-i386-pair/info.html
and here are the logs from the corresponding kernel build:
  http://www.chiark.greenend.org.uk/~xensrcts/logs/22171/build-i386-pvops/info.html

The failing host in that test is gall-mite.  When you look at the
logs, you want to start at around  Nov 28 16:42:20.112715
in serial-gall-mite.log.0.

I have attached my kconfig, although it can be found above too.

Please let me know what more information I should collect.

Ian.

View attachment "config-3.13.0-rc1+" of type "text/plain" (94602 bytes)