Re: [Qemu-devel] [PATCH v2 0/5] block: Avoid copy-on-read assertions

[Fam Zheng <famz@redhat.com>]

On Tue, 10/03 21:22, Eric Blake wrote:
> On 10/03/2017 09:16 PM, no-reply@patchew.org wrote:
> > Hi,
> > 
> > This series failed automatic build test. Please find the testing commands and
> > their output below. If you have docker installed, you can probably reproduce it
> > locally.
> > 
> 
> > 195         [not run] not suitable for this image format: raw
> > 197         - output mismatch (see 197.out.bad)
> > --- /tmp/qemu-test/src/tests/qemu-iotests/197.out	2017-10-04 01:52:59.000000000 +0000
> > +++ /tmp/qemu-test/build/tests/qemu-iotests/197.out.bad	2017-10-04 02:15:52.212004491 +0000
> > @@ -12,13 +12,18 @@
> >  128 KiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> >  read 0/0 bytes at offset 0
> >  0 bytes, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> > -read 2147483136/2147483136 bytes at offset 1024
> > -2 GiB, X ops; XX:XX:XX.X (XXX YYY/sec and XXX ops/sec)
> > +
> > +(process:16284): GLib-ERROR **: gmem.c:100: failed to allocate 2147483136 bytes
> 
> Okay, a test that requires a nearly-2G read in one operation is fringe,
> and I can see it choking 32-bit platforms rather easily.  How do we
> modify the test to not be so mean to memory-starved systems?  And why
> didn't patchew complain about this on v1, which had the same ~2G read?

I don't know. The whole system (Fedora VM) is dedicated to patchew test and no
concurrent task should be running. Maybe 2G is just in between the memory
watermarks.

Fam