exeseal: do not use F_SEAL_FUTURE_WRITE

[tomaszduda23]

Prior to kernel Linux 5.5 F_SEAL_FUTURE_WRITE has bug which maps memory as shared
between processes even if it is set as private.
torvalds/linux@05d3511

Accoring to https://man.archlinux.org/man/fcntl64.2.en F_SEAL_WRITE is enough.

Furthermore, trying to create new shared, writable memory-mappings via mmap(2)
will also fail with EPERM.

Using the F_ADD_SEALS operation to set the F_SEAL_WRITE seal fails with EBUSY
if any writable, shared mapping exists. Such mappings must be unmapped before
you can add this seal.

F_SEAL_FUTURE_WRITE make sens only if R/W shared mapping in one process should be
R/O in another process. This is not case for runc.

Fixes #4640

[cyphar]

Thanks for the patch and figuring out the issue! Honestly, I think it would be simpler just to drop F_SEAL_FUTURE_WRITE entirely.

To be honest, I didn't read the man page very closely when adding it (I thought it restricted dirty page writebacks or something like that), but it seems that it is basically a no-op with F_SEAL_WRITE aside from bugs like this.

[cyphar]

LGTM, thanks!

[kolyshkin]

A nit, but perhaps it's better to remove the code entirely, rather than commenting it out. At most, add a comment telling something like

// Don't add F_SEAL_FUTURE_WRITE, it is not needed (and is buggy for Linux < 5.5).

[kolyshkin]

lgtm

[kolyshkin]

• 1.2 backport: [1.2] exeseal: do not use F_SEAL_FUTURE_WRITE #4651