Thread (7 messages) 7 messages, 5 authors, 2017-07-27

Re: [BUG] nvme driver crash

From: Shaohua Li <shli@kernel.org>
Date: 2017-07-26 22:29:52

On Wed, Jul 26, 2017 at 12:12:21PM -0700, Omar Sandoval wrote:
On Wed, Jul 26, 2017 at 10:34:43AM -0700, Christoph Hellwig wrote:
quoted
On Tue, Jul 25, 2017 at 03:24:08PM -0700, Shaohua Li wrote:
quoted
Disable CONIFG_SMP, kernel crashes at boot time, here is the log.
I can reproduce the issue.  Unfortunately the addresss in the bug
doesn't make any sense to me when resolving it using gdb, as t just
points to the line where blk_mq_init_queue calls
blk_alloc_queue_node.

Can you check if you get better results in your build?
It's crashing on that line because ctrl->tagset is NULL, which is
because...

irq_create_affinity_masks() returns NULL on !CONFIG_SMP
-> pci_irq_get_affinity() returns NULL
-> blk_mq_pci_map_queues() returns -EINVAL
-> blk_mq_alloc_tag_set() returns -EINVAL
-> nvme_dev_add() doesn't set ctrl->tagset

The two-fold fix would be to make the nvme driver handle
blk_mq_alloc_tag_set() failing and to fall back to a dumb mapping in
blk_mq_pci_map_queues(), but I don't know what the best way to do those
is.
irq_create_affinity_masks returns null and prints error, could we fix that?
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help