[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1880d84c-b8fa-47ca-4fc7-191ccc6f976f@gmx.de>
Date: Mon, 26 Aug 2019 18:16:27 +0200
From: Matthias Andree <matthias.andree@....de>
To: Mika Westerberg <mika.westerberg@...ux.intel.com>,
Bjorn Helgaas <helgaas@...nel.org>
Cc: "Rafael J. Wysocki" <rjw@...ysocki.net>,
Len Brown <lenb@...nel.org>, Lukas Wunner <lukas@...ner.de>,
Keith Busch <keith.busch@...el.com>,
Alex Williamson <alex.williamson@...hat.com>,
Alexandru Gagniuc <mr.nuke.me@...il.com>,
Kai-Heng Feng <kai.heng.feng@...onical.com>,
Paul Menzel <pmenzel@...gen.mpg.de>,
Nicholas Johnson <nicholas.johnson-opensource@...look.com.au>,
Justin Forbes <jmforbes@...uxtx.org>,
linux-pci@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] PCI: Add missing link delays required by the PCIe spec
Am 26.08.19 um 16:42 schrieb Mika Westerberg:
> According to PCI FW 3.2 the OS is responsible for those delays if the
> platform firmware does not provide that _DSM. So simpler way would be
> always do the delays if the _DSM is not there.
Well, that is
* unless we trip over delays that make other parts of the kernel spin
crazy in unterminated loops because there's this mismatch between the
PME and the readyness world or whatever you analyzed my system PME
interrupt storms for.
Note we normally resume devices asynchronously these days unless someone
zeroes out /sys/power/pm_async.
* assuming we're not seeing complaints from I think it was Nicholas's
system with wakeup/resume deadlines.
Powered by blists - more mailing lists