On Sat, 30 Jan 2021 16:00:06 +0200 Neftin, Sasha wrote:
> On 1/30/2021 08:22, Jakub Kicinski wrote:
> > On Thu, 28 Jan 2021 13:38:48 -0800 Tony Nguyen wrote:
> >> From: Kai-Heng Feng <kai.heng.f...@canonical.com>
> >>
> >> Similar to commit 165ae7a8feb5 ("igb: Report speed and duplex as unknown
> >> when device is runtime suspended"), if we try to read speed and duplex
> >> sysfs while the device is runtime suspended, igc will complain and
> >> stops working:
> >
> >> The more generic approach will be wrap get_link_ksettings() with begin()
> >> and complete() callbacks, and calls runtime resume and runtime suspend
> >> routine respectively. However, igc is like igb, runtime resume routine
> >> uses rtnl_lock() which upper ethtool layer also uses.
> >>
> >> So to prevent a deadlock on rtnl, take a different approach, use
> >> pm_runtime_suspended() to avoid reading register while device is runtime
> >> suspended.
> >
> > Is someone working on the full fix to how PM operates?
> >
> > There is another rd32(IGC_STATUS) in this file which I don't think
> > is protected either.
>
> What is another rd32(IGC_STATUS) you meant? in igc_ethtool_get_regs?
Yes.
> While the device in D3 state there is no configuration space registers
> access.
That's to say similar stack trace will be generated to the one fixed
here, if someone runs ethtool -d, correct? I don't see anything
checking runtime there either.
To be clear I'm not asking for this to be addressed in this series.
Rather for a strong commitment that PM handling will be restructured.
It seems to me you should depend on refcounting / locking that the PM
subsystem does more rather than involving rtnl_lock.
