Date:   Fri, 12 Nov 2021 05:09:27 -0500
From:   "Michael S. Tsirkin" <mst@redhat.com>
To:     Gerd Hoffmann <kraxel@redhat.com>
Cc:     linux-pci@vger.kernel.org, Bjorn Helgaas <bhelgaas@google.com>,
        open list <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH] pciehp: fast unplug for virtual machines
Message-ID: <20211112050508-mutt-send-email-mst@kernel.org>
References: <20211111090225.946381-1-kraxel@redhat.com>
 <20211111115931-mutt-send-email-mst@kernel.org>
 <20211112095629.uoxfuhsvhicsdxgd@sirius.home.kraxel.org>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20211112095629.uoxfuhsvhicsdxgd@sirius.home.kraxel.org>
Precedence: bulk

On Fri, Nov 12, 2021 at 10:56:29AM +0100, Gerd Hoffmann wrote:
>   Hi,
> 
> > > This patch adds the fast_virtual_unplug module parameter to the
> > > pciehp driver.  When enabled (which is the default) the linux
> > > kernel will simply skip the delay for virtual pcie ports, which
> > > reduces the total time for the unplug operation from 6-7 seconds
> > > to 1-2 seconds.
> > 
> > BTW how come it's still taking seconds, not milliseconds?
> 
> I've tackled the 5 seconds only, biggest chunk and easy target because
> the only reason to have that is to allow operators press the attention
> button again to cancel, so the risk to break something here is rather
> low.
> 
> There are some more wait times elsewhere, to give the hardware the
> time needed when powering up/down slots, which sum up to roughly one
> second, and the time the driver needs to shutdown the device goes on
> top of that (typically not much).
> 
> take care,
>   Gerd

Probably also unnecessary for a virtual bridge. We'll need to fix these
up if we want native hotplug to be useful for workloads like kata -
these expect hotplug time in the milliseconds.

-- 
MST