BTRFS error (device nvme1n1p2): bdev /dev/nvme0n1p2 errs: wr 37868055...

[Chaitanya Kulkarni]

On 11/11/25 08:56, Chris Murphy wrote:
>
> On Mon, Nov 10, 2025, at 10:05 AM, Justin Piszcz wrote:
>> Hello,
>>
>> I am using an ASUS Pro WS W680-ACE motherboard with 2 x Samsung SSD
>> 990 PRO with Heatsink 4TB NVME SSDs with BTRFS R1.  When a BTRFS scrub
>> was kicked off this morning, suddenly BTRFS was noting errors for one
>> of the drives.  The system became unusable and I had to power cycle
>> and re-run the scrub and everything is now OK.  My question is what
>> would cause this?
> We'd have to see a complete dmesg at the time the problem occurred. If the same device holds system log files, seems like a pretty good chance none of it made it to persistent storage.
>
> All we know is btrfs sees a bunch of dropped reads, writes, and flush requests. So it's not a btrfs error per se, though it affects btrfs. The issue is with the NVMe drive, its firmware, or a kernel nvme driver bug, or some combination of those.
>
> --
> Chris Murphy
>
Isolate the problem between kernel and SSD FW by:-

1. run the same workload on different vendor SSDs.
2. run the same workload on qemu nvme emulation.

This will allow you to remove the SSD FW out of this question.

-ck