Re: [git pull] device mapper changes for 6.1

[Milan Broz <gmazyland@gmail.com>]

On 10/18/22 23:13, Mike Snitzer wrote:
...
>> In the absence of that kind of unification, just use 'errno'.
> 
> Mikulas touches on why why using errno isn't useful for DM. And for
> DM's device stacking it is hard to know which error spewed to dmesg
> (via DMERR, DMCRIT, DMINFO, etc) is relevant to a particular admin
> interface issuing the DM ioctl.
> 
> So the idea to send the (hopefully useful) error string back up to the
> relevant userspace consumer was one task that seemed needed (based on
> Milan Broz's various complaints against DM.. which sprang from your
> regular remainder that DM's version numbers are broken and need to be
> removed/replaced).

Well, when you mention my complaints, I think we are moving from one extreme
to another.

For the error reporting - we use errno values in libcryptsetup everywhere,
so if DM ioctls (through libdevmapper we use) returns proper errno, this is
the minimal solution that helps here.
The problem is that ioctl() are often just translated to -EINVAL.
(Or lost in libdevmapper compatibility layers.)

 From the dm-crypt/verity/integrity perspective, ENOMEM, ENODEV (bad block device),
ENOTSUP/ENOENT (for crypto algs not available), EIO for IO error,
EILSEQ for data integrity failure is the basic what we need.
(I perhaps forgot some, I can go through the code if you need it.)

As a bonus, if DM ioctl() returns fixed string that describes user-friendly error
(like: "keysize not compatible" or so) that's all we need
(ti->error strings are already in DM targets).

I have never asked for dynamically allocated error strings in kernel
and I do not know Mikulas' motivation to implement it.

Milan