Re: Preemptible idr_alloc() in QRTR code

On Tue, Jan 26, 2021 at 11:00:05AM -0600, Bjorn Andersson wrote:

On Tue 26 Jan 10:21 CST 2021, Mark Rutland wrote:

quoted

On Tue, Jan 26, 2021 at 02:58:33PM +0000, Matthew Wilcox wrote:

quoted

On Tue, Jan 26, 2021 at 10:47:34AM +0000, Mark Rutland wrote:

quoted

Hi,

When fuzzing arm64 with Syzkaller, I'm seeing some splats where
this_cpu_ptr() is used in the bowels of idr_alloc(), by way of
radix_tree_node_alloc(), in a preemptible context:

I sent a patch to fix this last June.  The maintainer seems to be
under the impression that I care an awful lot more about their
code than I do.

https://lore.kernel.org/netdev/20200605120037.17427-1-willy@infradead.org/ (local)

Ah; I hadn't spotted the (glaringly obvious) GFP_ATOMIC abuse, thanks
for the pointer, and sorry for the noise.

I'm afraid this isn't as obvious to me as it is to you. Are you saying
that one must not use GFP_ATOMIC in non-atomic contexts?

That said, glancing at the code I'm puzzled to why it would use
GFP_ATOMIC.

I'm also not entirely sure about the legitimacy of GFP_ATOMIC outside of
atomic contexts -- I couldn't spot any documentation saying that wasn't
legitimate, but Matthew's commit message implies so, and it sticks out
as odd.

quoted

It looks like Eric was after a fix that trivially backported to v4.7
(and hence couldn't rely on xarray) but instead it just got left broken
for months. :/

Bjorn, is this something you care about? You seem to have the most
commits to the file, and otherwise the official maintainer is Dave
Miller per get_maintainer.pl.

I certainly care about qrtr working and remember glancing at Matthew's
patch, but seems like I never found time to properly review it.

quoted

It is very tempting to make the config option depend on BROKEN...

I hear you and that would be bad, so I'll make sure to take a proper
look at this and Matthew's patch.

Thanks! I'm happy to try/test patches if that's any help. My main
concern here is that this can be triggered on arbitrary platforms so
long as the driver is built in (e.g. my Syzkaller instance is hitting
this within a VM).

Thanks,
Mark.