uffd: fixed for exiting exiting program as user

[nithin-jaikar]

While Dumping the process, it fails and exits as a user because of kerndat_uffd() returns -1 and due permission error
uffd = syscall(SYS_userfaultfd, flags); this function fails inturn kerndat_uffd() fails.
In the change it ignores the permission error and proceeds further.

[xemul]

@nithin-jaikar rework the PR, leave only the EPERM checking of uffd opening and write a proper patch comment.

[xemul]

@nithin-jaikar , don't add 5th commit, please. Remove all commits and leave only one.

[xemul]

The patch (not PR, but the very commit) must have a descriptive comment

[xemul]

@nithin-jaikar , you've somehow pulled in the Adrian's patch, then merged yours with his and then added the fix I requested as another commit. Please, be more attentive and make a proper PR.

[xemul]

Ok, @nithin-jaikar , now let's add the comment to the commit (the text that accompanies the pull-request can be just copied)

[adrianreber]

From the PR description it is not clear to me what this tries to solve. Some more details how this could happen would be interesting.

[xemul]

@adrianreber , this is the part of "dump shared memory as user" task.

The "criu dump" used to work from non-root some time ago, but then this mode testing stopped and the functionality got broken. On of the found problems is that non-root is not allowed to open uffd, so criu cannot even check the kerndat stuff.

[adrianreber]

Ah, thanks. I saw similar errors when doing the non-root changes in #1155

Anyway, that information would be good to have in the commit and/or PR.

[xemul]

Whee, comment commit message is mastered!
@nithin-jaikar , now please read https://criu.org/How_to_submit_patches and make the commit message look the way it must. E.g. like this

kerndat: Handle non-root mode when checking uffd

When criu is run as user it fails and exits because of kerndat_uffd() returning -1.
This, in turn, happens after uffd = syscall(SYS_userfaultfd, flags); which only works
for root.

In the change it ignores the permission error and proceeds further just like it's done
for e.g. pagemap checking.

Signed-off-by: Nithin Jaikar J <[email protected]>

[rst0git]

LGTM

[Snorch]

#define EPERM 1

Should not we differentiate errno == EPERM from syscall SYS_userfaultfd and later errors in uffd_open() ?

[Snorch]

I still think that my objection is valid, so #1493