On Wed, May 20, 2020 at 05:51:27AM -0700, Eric Dumazet wrote:
>
> On 5/19/20 11:42 PM, Ahmed S. Darwish wrote:
> > Hello Eric,
> >
> > On Tue, May 19, 2020 at 07:01:38PM -0700, Eric Dumazet wrote:
> >>
> >> On 5/19/20 2:45 PM, Ahmed S. Darwish wrote:
> >>> Sequence counters write paths are critical sections that must never be
> >>> preempted, and blocking, even for CONFIG_PREEMPTION=n, is not allowed.
> >>>
> >>> Commit 5dbe7c178d3f ("net: fix kernel deadlock with interface rename and
> >>> netdev name retrieval.") handled a deadlock, observed with
> >>> CONFIG_PREEMPTION=n, where the devnet_rename seqcount read side was
> >>> infinitely spinning: it got scheduled after the seqcount write side
> >>> blocked inside its own critical section.
> >>>
> >>> To fix that deadlock, among other issues, the commit added a
> >>> cond_resched() inside the read side section. While this will get the
> >>> non-preemptible kernel eventually unstuck, the seqcount reader is fully
> >>> exhausting its slice just spinning -- until TIF_NEED_RESCHED is set.
> >>>
> >>> The fix is also still broken: if the seqcount reader belongs to a
> >>> real-time scheduling policy, it can spin forever and the kernel will
> >>> livelock.
> >>>
> >>> Disabling preemption over the seqcount write side critical section will
> >>> not work: inside it are a number of GFP_KERNEL allocations and mutex
> >>> locking through the drivers/base/ :: device_rename() call chain.
> >>>
> >>> From all the above, replace the seqcount with a rwsem.
> >>>
> >>> Fixes: 5dbe7c178d3f (net: fix kernel deadlock with interface rename and
> >>> netdev name retrieval.)
> >>> Fixes: 30e6c9fa93cf (net: devnet_rename_seq should be a seqcount)
> >>> Fixes: c91f6df2db49 (sockopt: Change getsockopt() of SO_BINDTODEVICE to
> >>> return an interface name)
> >>> Cc: <sta...@vger.kernel.org>
> >>> Signed-off-by: Ahmed S. Darwish <a.darw...@linutronix.de>
> >>> Reviewed-by: Sebastian Andrzej Siewior <bige...@linutronix.de>
> >>> ---
> >>> net/core/dev.c | 30 ++++++++++++------------------
> >>> 1 file changed, 12 insertions(+), 18 deletions(-)
> >>>
> >>
> >> Seems fine to me, assuming rwsem prevent starvation of the writer.
> >>
> >
> > Thanks for the review.
> >
> > AFAIK, due to 5cfd92e12e13 ("locking/rwsem: Adaptive disabling of reader
> > optimistic spinning"), using a rwsem shouldn't lead to writer starvation
> > in the contended case.
>
> Hmm this was in linux-5.3, so very recent stuff.
>
> Has this patch been backported to stable releases ?
>
> With all the Fixes: tags you added, stable teams will backport this
> networking patch to all stable versions.
>
> Do we have a way to tune a dedicare rwsem to 'give preference to the
> (unique in this case) writer" over a myriad of potential readers ?
>
I was wrong in referencing the commit 5cfd92e12e13 above.
Before and after that commit, once a rwsem writer is blocking, all
subsequent readers will block until that writer makes progress.
Given that behavior, and that the read section is already quite short, I
don't think there's any danger incurred on writers here.
(a v2 will be sent shortly, fixing the error found Dan/kbuild-bot.)
Thanks,
--
Ahmed S. Darwish
Linutronix GmbH
