Re: dm-verity design question

Hi Jasper,

if it is question about kernel code, the best is to ask on [email protected] list
(you can cc cryptsetup list too).

Not all device-mapper kernel developers are here (dm-devel is official list
for kernel device-mapper subsystem; this one mainly for userspace cryptsetup tools).

Milan


On 12/04/2022 11:38, Jasper Surmont wrote:
> Hey,
> 
> After going through the dm-verity source code for a while I think I
> kind of figured out the general design; on a read:
> 1. Multiple possible problems get checked like alignment, out of range
> etc. Also variables get set based on the request
> 2. The request is prefetched to the cache
> 3. The actual integrity check is done when the bio ends (bi_end_io
> points to verity_end_io which submits the work to verify digest etc
> 
> If this is correct, I have 2 questions:
> 1. What is the main benefit of prefetching here? We know the remapped
> request is going to be executed very soon, so I don't really see why
> we prefetch.
> 2. In verity_end_io(...), we only submit the work to verify the digest
> if FEC is enabled. Why? Can't we check integrity without using FEC?
> 
> I hope my questions are clear.
> 
> Thanks a lot!
> 
> Sincerely, Jasper Surmont
>