Re: nvme blk_update_request IO error is seen on stable kernel 5.4.41.

[Dakshaja Uppalapati <dakshaja@chelsio.com>]

[-- Attachment #1: Type: text/plain, Size: 2260 bytes --]

On Monday, June 06/01/20, 2020 at 18:27:50 +0200, Greg KH wrote:
> On Mon, Jun 01, 2020 at 09:51:44PM +0530, Dakshaja Uppalapati wrote:
> > On Thursday, May 05/28/20, 2020 at 10:34:03 +0200, Greg KH wrote:
> > > On Thu, May 28, 2020 at 01:14:31PM +0530, Dakshaja Uppalapati wrote:
> > > > On Tuesday, May 05/26/20, 2020 at 12:25:42 +0200, Greg KH wrote:
> > > > > On Thu, May 21, 2020 at 07:36:43PM +0530, Dakshaja Uppalapati wrote:
> > > > > > Hi all,
> > > > > > 
> > > > > > Issue which is reported in https://lore.kernel.org/linux-nvme/CH2PR12MB40050ACF
> > > > > > 2C0DC7439355ED3FDD270@CH2PR12MB4005.namprd12.prod.outlook.com/T/#r8cfc80b26f0cd
> > > > > > 1cde41879a68fd6a71186e9594c is also seen on stable kernel 5.4.41. 
> > > > > 
> > > > > What issue is that?  Your url is wrapped and can not work here :(
> > > > 
> > > > Sorry for that, when I tried to format the disk discovered from target machine
> > > > the below error is seen in dmesg.
> > > > 
> > > > dmesg:
> > > > 	[ 1844.868480] blk_update_request: I/O error, dev nvme0c0n1, sector 0 
> > > > 	op 0x3:(DISCARD) flags 0x4000800 phys_seg 1 prio class 0
> > > > 
> > > > The above issue is seen from kernel-5.5-rc1 onwards.
> > > > 
> > > > > 
> > > > > > In upstream issue is fixed with commit b716e6889c95f64b.
> > > > > 
> > > > > Is this a regression or support for something new that has never worked
> > > > > before?
> > > > > 
> > > > 
> > > > This is a regression, bisects points to the commit 530436c4 and fixed with
> > > > commit b716e688 in upstream.
> > > > 
> > > > Now same issue is seen with stable kernel-5.4.41, 530436c4 is part of it.
> > > 
> > > So why don't we just revert 530436c45ef2 ("nvme: Discard workaround for
> > > non-conformant devices") from the stable trees?  Will that fix the issue
> > > for you instead of the much-larger set of backports you are proposing?
> > > 
> > > Also, is this an issue for you in the 4.19 releases?  The above
> > > mentioned patch showed up in 4.19.92 and 5.4.7.
> > > 
> > 
> > Yes, on 4.19 stable kernel too issue is seen. By reverting 530436c45ef2 issue
> > is not seen on both 4.19 and 5.4 stable kernels. Do you want me to send the
> > reverted patch?
> 
> Yes please.

Attached the reverted patch.PFA.

Thanks,
Dakshaja

[-- Attachment #2: revert_530436c4.patch --]
[-- Type: text/plain, Size: 1212 bytes --]

diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
index f0e0af3aa..d658c5093 100644
--- a/drivers/nvme/host/core.c
+++ b/drivers/nvme/host/core.c
@@ -612,14 +612,8 @@ static blk_status_t nvme_setup_discard(struct nvme_ns *ns, struct request *req,
 	struct nvme_dsm_range *range;
 	struct bio *bio;
 
-	/*
-	 * Some devices do not consider the DSM 'Number of Ranges' field when
-	 * determining how much data to DMA. Always allocate memory for maximum
-	 * number of segments to prevent device reading beyond end of buffer.
-	 */
-	static const size_t alloc_size = sizeof(*range) * NVME_DSM_MAX_RANGES;
-
-	range = kzalloc(alloc_size, GFP_ATOMIC | __GFP_NOWARN);
+	range = kmalloc_array(segments, sizeof(*range),
+				GFP_ATOMIC | __GFP_NOWARN);
 	if (!range) {
 		/*
 		 * If we fail allocation our range, fallback to the controller
@@ -659,7 +653,7 @@ static blk_status_t nvme_setup_discard(struct nvme_ns *ns, struct request *req,
 
 	req->special_vec.bv_page = virt_to_page(range);
 	req->special_vec.bv_offset = offset_in_page(range);
-	req->special_vec.bv_len = alloc_size;
+	req->special_vec.bv_len = sizeof(*range) * segments;
 	req->rq_flags |= RQF_SPECIAL_PAYLOAD;
 
 	return BLK_STS_OK;