[Qemu-devel] [ANNOUNCE] QEMU 0.14.0-rc2 is available

[Qemu-devel] [ANNOUNCE] QEMU 0.14.0-rc2 is available

[Anthony Liguori]

The QEMU team is pleased to announce the availability of the 0.14.0-rc2 
release.   This is the third release candidate for the 0.14.0 release. 
A detailed changelog is at the end of this message.

Download instructions are available at http://wiki.qemu.org/Download

You can help test this release by signing up on the new Testing wiki 
page at http://wiki.qemu.org/Planning/0.14/Testing

An in progress detailed change log is available at 
http://wiki.qemu.org/Changelog/0.14

On behalf of the QEMU team, I'd like to thank everyone who contributed 
to make this release happen!

Since v0.14.0-rc1:

  - slirp: fix buffer overrun (Bruce Rogers)
  - correctly check ppr priority during interrupt injection] (Gleb  Natapov)
  - qcow2: Fix order in L2 table COW (Kevin Wolf)
  - qemu-img: Improve error messages for failed bdrv_open (Kevin Wolf)
  - qed: Report error for unsupported features (Kevin Wolf)
  - qcow2: Report error for version > 2 (Kevin Wolf)
  - qerror: Add QERR_UNKNOWN_BLOCK_FORMAT_FEATURE (Kevin Wolf)
  - qcow2: Fix error handling for reading compressed clusters (Kevin Wolf)
  - qcow2: Fix error handling for immediate backing file read failure 
(Kevin Wolf
  - QCOW2: bug fix - read base image beyond its size (Chunqiang Tang)
  - Change snapshot_blkdev hmp to use correct argument type for device 
(Jes Sorenson)
  - linux-user: Fix possible realloc memory leak (Stefan Weil)
  - linux-user: Fix possible realloc memory leak (Stefan Weil)
  - linux-user: fix for loopmount ioctl (Martin Mohring)
  - blockdev: Plug memory leak in drive_init() error paths (Markus 
Armbruster)
  - blockdev: Plug memory leak in drive_uninit() (Markus Armbruster)
  - x86: Fix MCA broadcast parameters for TCG case (Jan Kiszka)
  - qemu-timer: Fix compilation of new timer code for w32, w64 (Stefan Weil)


Regards,

Anthony Liguori

Re: [Qemu-devel] [ANNOUNCE] QEMU 0.14.0-rc2 is available

[Rick Vernam]

Using 0.14.0-rc2 it looks like a SPICE pid file is left over in /dev/shm, ie 
spice.10005, after qemu closes.

Re: [Qemu-devel] [ANNOUNCE] QEMU 0.14.0-rc2 is available

[Anthony Liguori]

On 02/16/2011 10:28 AM, Rick Vernam wrote:
> Using 0.14.0-rc2 it looks like a SPICE pid file is left over in /dev/shm, ie
> spice.10005, after qemu closes.
>    

I think that's coming from libspice, not QEMU.

Regards,

Anthony Liguori

Re: [Qemu-devel] [ANNOUNCE] QEMU 0.14.0-rc2 is available

[Gerd Hoffmann]

On 02/16/11 17:37, Anthony Liguori wrote:
> On 02/16/2011 10:28 AM, Rick Vernam wrote:
>> Using 0.14.0-rc2 it looks like a SPICE pid file is left over in
>> /dev/shm, ie
>> spice.10005, after qemu closes.
>
> I think that's coming from libspice, not QEMU.

Indeed.  It isn't a pidfile, it is a shared memory segment where spice 
stores statistics.  spice registers a cleanup handler via atexit() which 
removes this, so you should not see this unless qemu crashes and thus 
the atexit() handlers are not called.

cheers,
   Gerd

Re: [Qemu-devel] [ANNOUNCE] QEMU 0.14.0-rc2 is available

[Anthony Liguori]

On 02/16/2011 10:52 AM, Gerd Hoffmann wrote:
> On 02/16/11 17:37, Anthony Liguori wrote:
>> On 02/16/2011 10:28 AM, Rick Vernam wrote:
>>> Using 0.14.0-rc2 it looks like a SPICE pid file is left over in
>>> /dev/shm, ie
>>> spice.10005, after qemu closes.
>>
>> I think that's coming from libspice, not QEMU.
>
> Indeed.  It isn't a pidfile, it is a shared memory segment where spice 
> stores statistics.  spice registers a cleanup handler via atexit() 
> which removes this

fork() with an inherited pipe and then waiting for EOF on the pipe() is 
a bit more robust of a mechanism to handle this.

Regards,

Anthony Liguori

> , so you should not see this unless qemu crashes and thus the atexit() 
> handlers are not called.
>
> cheers,
>   Gerd
>
>

Re: [Qemu-devel] [ANNOUNCE] QEMU 0.14.0-rc2 is available

[Gerd Hoffmann]

   Hi,

>> Indeed. It isn't a pidfile, it is a shared memory segment where spice
>> stores statistics. spice registers a cleanup handler via atexit()
>> which removes this
>
> fork() with an inherited pipe and then waiting for EOF on the pipe() is
> a bit more robust of a mechanism to handle this.

Feels a bit like overkill in that case though.  The extra process 
doesn't come for free.  We are talking about a shared memory segment 
with a single page only.  It is leaked only on crashes.  It can 
trivially be cleaned up manually if needed, just rm the file.

cheers,
   Gerd

Re: [Qemu-devel] [ANNOUNCE] QEMU 0.14.0-rc2 is available

[Gerd Hoffmann]

On 02/16/11 17:52, Gerd Hoffmann wrote:
> On 02/16/11 17:37, Anthony Liguori wrote:
>> On 02/16/2011 10:28 AM, Rick Vernam wrote:
>>> Using 0.14.0-rc2 it looks like a SPICE pid file is left over in
>>> /dev/shm, ie
>>> spice.10005, after qemu closes.
>>
>> I think that's coming from libspice, not QEMU.
>
> Indeed. It isn't a pidfile, it is a shared memory segment where spice
> stores statistics. spice registers a cleanup handler via atexit() which
> removes this, so you should not see this unless qemu crashes and thus
> the atexit() handlers are not called.

Oh, and for completeness: spice client creates such files with stats too.

cheers,
   Gerd