Consider getting rid of u2mfn module and use /proc/PID/pagemap

[marmarek]

Starting with 2.6.25 Linux kernel does provide an interface to get MFN out of process virtual address. Which means we could replace custom u2mfn with that. While the change theoretically is small, this would mean we could use stock kernel, without any custom modules in the VM.

[marmarek]

Documentation: http://www.mjmwired.net/kernel/Documentation/vm/pagemap.txt

[marmarek]

Documentation: http://www.mjmwired.net/kernel/Documentation/vm/pagemap.txt

[marmarek]

any idea what config options pagemap is tied to? I can't see it.

Looks like CONFIG_PROC_PAGE_MONITOR

I also imagine it's somewhat harmful to ASLR,

Process owner can't access it - it requires CAP_SYS_ADMIN. From the above documentation:

Since Linux 4.0 only users with the CAP_SYS_ADMIN capability can get PFNs.
In 4.0 and 4.1 opens by unprivileged fail with -EPERM. Starting from
4.2 the PFN field is zeroed if the user does not have CAP_SYS_ADMIN.
Reason: information about PFNs helps in exploiting Rowhammer vulnerability.

Anyway, it looks like /proc/$pid/pagemap doesn't contain enough information to really replace u2mfn. It gives PFNs (guest-specific pseudo-physical frame number), but we need MFNs (machine physical frame number).

[marmarek]

any idea what config options pagemap is tied to? I can't see it.

Looks like CONFIG_PROC_PAGE_MONITOR

I also imagine it's somewhat harmful to ASLR,

Process owner can't access it - it requires CAP_SYS_ADMIN. From the above documentation:

Since Linux 4.0 only users with the CAP_SYS_ADMIN capability can get PFNs.
In 4.0 and 4.1 opens by unprivileged fail with -EPERM. Starting from
4.2 the PFN field is zeroed if the user does not have CAP_SYS_ADMIN.
Reason: information about PFNs helps in exploiting Rowhammer vulnerability.

Anyway, it looks like /proc/$pid/pagemap doesn't contain enough information to really replace u2mfn. It gives PFNs (guest-specific pseudo-physical frame number), but we need MFNs (machine physical frame number).