Re: [PATCH v6 0/3] ublk: introduce UBLK_CMD_TRY_STOP_DEV

[Jens Axboe <axboe@kernel.dk>]

On 1/12/26 9:28 AM, Caleb Sander Mateos wrote:
> On Mon, Jan 12, 2026 at 8:22?AM Jens Axboe <axboe@kernel.dk> wrote:
>>
>> Please rebase this on the current for-7.0/block tree. It doesn't
>> apply at all, and hunks like:
>>
>> @@ -311,6 +312,12 @@
>>   */
>>  #define UBLK_F_BUF_REG_OFF_DAEMON (1ULL << 14)
>>
>> +/*
>> + * The device supports the UBLK_CMD_TRY_STOP_DEV command, which
>> + * allows stopping the device only if there are no openers.
>> + */
>> +#define UBLK_F_SAFE_STOP_DEV   (1ULL << 17)
>> +
>>  /* device state */
>>  #define UBLK_S_DEV_DEAD        0
>>  #define UBLK_S_DEV_LIVE        1
>>
>> is a clear sign that your way off base at this point. Why else would
>> STOP_DEV be 1 << 17, with the previous one at 1 << 14?
> 
> This is due to being developed in parallel with other ublk patch sets
> which are using feature flags 15 (UBLK_F_BATCH_IO) and 16
> (UBLK_F_INTEGRITY). UBLK_F_BATCH_IO is still in development and
> UBLK_F_INTEGRITY was just applied. So I don't think it's fair to blame
> Yoav for not having rebased on commits that didn't exist yet when this
> version of the patch series was sent out :) Should be a trivial
> rebase, though, since the value of 17 was already chosen to avoid
> conflicting with the other patch sets.

At the time v6 was posted, it did not apply to the current branches. So
regardless of how it was developed, that's not a very useful approach as
it means I have to hand apply it. Which isn't too difficult for patches
1-2. But now we have the integrity bits in, which means the selftest
bits take more work. At that point I just threw my hands up.

Hence please repost against the current tree.

-- 
Jens Axboe