RE: [PATCH v7 00/16] Intel IPU3 ImgU patchset

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi Jacopo,

> Subject: Re: [PATCH v7 00/16] Intel IPU3 ImgU patchset
> 
> Hello Raj,
> 
> On Wed, Jan 09, 2019 at 05:00:21PM +0000, Mani, Rajmohan wrote:
> > Hi Laurent, Tomasz, Jacopo,
> >
> > > Subject: Re: [PATCH v7 00/16] Intel IPU3 ImgU patchset
> > >
> > > Hello,
> > >
> > > On Tue, Jan 08, 2019 at 03:54:34PM +0900, Tomasz Figa wrote:
> > > > Hi Raj, Yong, Bingbu, Tianshu,
> > > >
> > > > On Fri, Dec 21, 2018 at 12:04 PM Tomasz Figa <tfiga@xxxxxxxxxxxx>
> wrote:
> > > > >
> > > > > On Fri, Dec 21, 2018 at 7:24 AM Laurent Pinchart
> > > > > <laurent.pinchart@xxxxxxxxxxxxxxxx> wrote:
> > > > > >
> > > > > > Hellon
> > > > > >
> > > > > > On Sunday, 16 December 2018 09:26:18 EET Laurent Pinchart wrote:
> > > > > > > Hello Yong,
> > > > > > >
> > > > > > > Could you please have a look at the crash reported below ?
> > > > > >
> > > > > > A bit more information to help you debugging this. I've
> > > > > > enabled KASAN in the kernel configuration, and get the
> > > > > > following use-after-free
> > > reports.
> > >
> > > I tested as well using the ipu-process.sh script shared by Laurent,
> > > with the following command line:
> > > ./ipu3-process.sh --out 2560x1920 --vf 1920x1080
> > > frame-2592x1944.cio2
> > >
> > > and I got a very similar trace available at:
> > > https://paste.debian.net/hidden/5855e15a/
> > >
> > > Please note I have been able to process a set of images (with KASAN
> > > enabled the machine does not freeze) but the kernel log gets flooded
> > > and it is not possible to process any other frame after this.
> > >
> > > The issue is currently quite annoying and it's a blocker for
> > > libcamera development on IPU3. Please let me know if I can support with
> more testing.
> > >
> > > Thanks
> > >    j
> > >
> > > > > >
> > > > > > [  166.332920]
> > > > > >
> > >
> ================================================================
> > > ==
> > > > > > [  166.332937] BUG: KASAN: use-after-free in
> > > > > > __cached_rbnode_delete_update+0x36/0x202
> > > > > > [  166.332944] Read of size 8 at addr ffff888133823718 by task
> > > > > > yavta/1305
> > > > > >
> > > > > > [  166.332955] CPU: 3 PID: 1305 Comm: yavta Tainted: G         C
> 4.20.0-
> > > rc6+ #3
> > > > > > [  166.332958] Hardware name: HP Soraka/Soraka, BIOS
> > > > > > 08/30/2018 [ 166.332959] Call Trace:
> > > > > > [  166.332967]  dump_stack+0x5b/0x81 [  166.332974]
> > > > > > print_address_description+0x65/0x227
> > > > > > [  166.332979]  ? __cached_rbnode_delete_update+0x36/0x202
> > > > > > [  166.332983]  kasan_report+0x247/0x285 [  166.332989]
> > > > > > __cached_rbnode_delete_update+0x36/0x202
> > > > > > [  166.332995]  private_free_iova+0x57/0x6d [  166.332999]
> > > > > > __free_iova+0x23/0x31 [  166.333011]
> > > > > > ipu3_dmamap_free+0x118/0x1d6 [ipu3_imgu]
> > > > >
> > > > > Thanks Laurent, I think this is a very good hint. It looks like
> > > > > we're basically freeing and already freed IOVA and corrupting
> > > > > some allocator state?
> > > >
> > > > Did you have any luck in reproducing and fixing this double free issue?
> > > >
> >
> > This issue is either hard to reproduce or comes with different
> > signatures with the updated yavta (that now supports meta output) with
> > the 4.4 kernel that I have been using.
> > I am switching to 4.20-rc6 for better reproducibility.
> > Enabling KASAN also results in storage space issues on my Chrome device.
> > Will enable this just for ImgU to get ahead and get back with more updates.
> >
> 
> Thanks for testing this.
> 
> For your informations I'm using the following branch, from Sakari's
> tree: git://linuxtv.org/sailus/media_tree.git ipu3
> 
> Although it appears that the media tree master branch has everything that is
> there, with a few additional patches on top. I should move to use media tree
> master as well...
> 
> I have here attached 2 configuration files for v4.20-rc5 I am using on Soraka, in
> case they might help you. One has KASAN enabled with an increased kernel
> log size, the other one is the one we use for daily development.

I think I am missing a trick here to override the default chrome os kernel
config with the one that you supplied.

In particular I am looking for steps to build the upstream kernel within chrome os
build environment using your config, so I can update my Soraka device.

> 
> Also, please make sure to use (the most) recent media-ctl and yavta utilities, as
> the ones provided by most distros are usually not recent enough to work with
> IPU3, but I'm sure you know that already ;)

Ack

> 
> Thanks
>   j
> 

[snip]




[Index of Archives]     [Linux Input]     [Video for Linux]     [Gstreamer Embedded]     [Mplayer Users]     [Linux USB Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [Yosemite Backpacking]

  Powered by Linux