[PATCH for-stable nvmet 0/6] nvme: Fix for blk_update_request IO error.

[Dakshaja Uppalapati]

On Wednesday, June 06/17/20, 2020 at 07:15:41 -0700, Keith Busch wrote:
> On Thu, Jun 11, 2020 at 09:23:33PM +0530, Dakshaja Uppalapati wrote:
> > The below error is seen in dmesg, while formatting the disks discovered on host.
> > 
> > dmesg:
> >         [  636.733374] blk_update_request: I/O error, dev nvme4n1, sector 0 op 0x3:(DISCARD) flags 0x800 phys_seg 1 prio class 0
> > 
> > Patch 6 fixes it and there are 5 other dependent patches that also need to be 
> > pulled from upstream to stable, 5.4 and 4.19 branches.
> > 
> > Patch 1 dependent patch
> > 
> > Patch 2 dependent patch
> > 
> > Patch 3 dependent patch
> > 
> > Patch 4 dependent patch
> > 
> > Patch 5 dependent patch
> > 
> > Patch 6 fix patch
> 
> 1. You need to copy the linux-nvme mainling list for linux nvme kernel patches.
> 
> 2. If you're sending someone else's patch, the patch is supposed to have
> the From: tag so the author is appropriately identified.
> 
> 3. Stable patches must referece the upstream commit ID.
> 
> As for this particular issue, while stable patches are required to
> reference an upstream commit, you don't need to bring in dependent
> patches. You are allowed to write an equivalent fix specific to the
> stable branch so that stable doesn't need to take a bunch of unrelated
> changes. For example, it looks like this particular isssue can be fixed
> with the following simple stable patch:
>

Hi keith,

Thanks for the review.

I initially tried pushing only the fix + required portion of the dependent 
patches(https://www.spinics.net/lists/stable/msg387744.html) but as that 
approach is discouraged in stable tree, I submitted all the patches as it is.

Here are the ways to fix the issue in stable tree:

•  push fix + all dependent patches
•  push fix + custom patch of dependent patches
•  revert the culprit patch.

Please let me know how this issue can be resolved in stable tree.

Thanks,
Dakshaja