Re: [PATCH v3 5/5] block: use chunk_sectors when evaluating stacked atomic write limits

[John Garry <john.g.garry@oracle.com>]

On 03/07/2025 16:36, Mikulas Patocka wrote:
>> I suppose theoretically it could happen, and I'm happy to change.
>>
>> However there seems to be precedent in assuming it won't:
>>
>> - in stripe_op_hints(), we hold chunk_size in an unsigned int
>> - in raid0_set_limits(), we hold mddev->chunk_sectors << 9 in lim.io_min,
>> which is an unsigned int type.
>>
>> Please let me know your thoughts on also changing these sort of instances. Is
>> it realistic to expect chunk_bytes > UINT_MAX?
>>
>> Thanks,
>> John
> dm-stripe can be created with a stripe size that is more than 0xffffffff
> bytes.
> 
> Though, the integer overflow already exists in the existing dm-stripe
> target:
> static void stripe_io_hints(struct dm_target *ti,
>                              struct queue_limits *limits)
> {
>          struct stripe_c *sc = ti->private;
>          unsigned int chunk_size = sc->chunk_size << SECTOR_SHIFT;
> 
>          limits->io_min = chunk_size;
>          limits->io_opt = chunk_size * sc->stripes;
> }
> What should we set there as io_min and io_opt if sc->chunk_size <<
> SECTOR_SHIFT overflows?


> Should we set nothing?

For io_min/opt, maybe reduce to a factor of the stripe size / width (and 
which fits in a unsigned int).

I am not sure if it is even sane to have such huge values in io_min and 
the bottom disk io_min should be used directly instead.

Martin Petersen might have a better idea.. he added those sysfs files :)

Thanks,
John