RE: [PATCH 1/2] PCI: Add pci_set_vpd_timeout() to set VPD access timeout

I had old Marvell Sky2 hardware that had slow flash.

-----Original Message-----
From: Andrew Donnellan [mailto:andrew.donnellan@au1.ibm.com]=20
Sent: Monday, November 21, 2016 4:16 PM
To: Bjorn Helgaas <helgaas@kernel.org>; Matthew R. Ochs <mrochs@linux.vnet.=
ibm.com>
Cc: linux-pci@vger.kernel.org; Frederic Barrat <redacted>=
; Uma Krishnan [off-list ref]; Ian Munsie <imunsie@au1.ibm.co=
m>; Bjorn Helgaas [off-list ref]; linuxppc-dev@lists.ozlabs.org; Ste=
phen Hemminger [off-list ref]
Subject: Re: [PATCH 1/2] PCI: Add pci_set_vpd_timeout() to set VPD access t=
imeout

On 22/11/16 09:05, Bjorn Helgaas wrote:

Hi Matthew,

On Mon, Nov 21, 2016 at 03:09:49PM -0600, Matthew R. Ochs wrote:

quoted

The PCI core uses a fixed 50ms timeout when waiting for VPD accesses=20
to complete. When an access does not complete within this period, a=20
warning is logged and an error returned to the caller.

While this default timeout is valid for most hardware, some devices=20
can experience longer access delays under certain circumstances. For=20
example, one of the IBM CXL Flash devices can take up to ~120ms in a=20
worst-case scenario. These types of devices can benefit from an=20
extended timeout that is specific to their hardware constraints.

To support per-device VPD access timeouts, pci_set_vpd_timeout() is=20
added as an exported service. PCI devices will continue to default=20
with the 50ms timeout and use a per-device timeout when a driver calls t=

his new service.

Can you include a pointer to something in the spec that's behind the=20
default 50ms timeout, or did somebody just pull that number out of the=20
air?

It looks like Stephen Hemminger added the 50ms timeout in 1120f8b8169f, whi=
ch seems to indicate that 50ms was chosen because it's longer than the 13ms=
 per word that was measured on one device.

--=20
Andrew Donnellan              OzLabs, ADL Canberra
andrew.donnellan@au1.ibm.com  IBM Australia Limited