]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/commit
UBUNTU: SAUCE: (no-up) NVMe: only setup MSIX once
authorDan Streetman <dan.streetman@canonical.com>
Thu, 8 Dec 2016 18:31:15 +0000 (13:31 -0500)
committerBrad Figg <brad.figg@canonical.com>
Thu, 8 Dec 2016 18:58:04 +0000 (10:58 -0800)
commit96fce9e4025b96b08bfe5196d3380ab9215cb64b
tree6905e9372b4c92e1930b6794c4ec5ab845f24aa2
parent764d47217b1f3881600e11c08f109b177e521b15
UBUNTU: SAUCE: (no-up) NVMe: only setup MSIX once

BugLink: http://bugs.launchpad.net/bugs/1648449
The NVMe driver currently configures a single MSI/MSIX interrupt for its
admin queue, then later after initializing the NVMe controller and
detecting the number of queues the controller can handle, the driver
releases its MSI/MSIX interrupt and immediately re-configures enough
MSI/MSIX interrupts for all the queues.  However, on Amazon AWS instances
with passthrough NVMe drives, this MSI/MSIX release/request causes
failures of some of the interrupts, resulting in some (usually all but one
or two but sometimes none) of the NVMe drive(s) failing to initialize.

This combines the MSI/MSIX configurations, so all the MSI/MSIX configuration
is done once.  This avoids the interrupt failures, but it also results in
more MSI/MSIX interrupts being configured than are needed.

Signed-off-by: Dan Streetman <dan.streetman@canonical.com>
Signed-off-by: Brad Figg <brad.figg@canonical.com>
drivers/nvme/host/pci.c