Closed thomasdangl closed 3 years ago
@adlazar can I have your input on this one ?
I'm not sure if QEMU fails if the ioctl fails when using the old value
On a second thought, if QEMU is not started with mem-introspection stuff, it should not fail (I guess :) ).
On a second thought, if QEMU is not started with mem-introspection stuff, it should not fail (I guess :) ).
Yup, when not using mem-introspection it appears to work fine. I will prepare a pull request for QEMU anyway if that is fine with you.
I will prepare a pull request for QEMU anyway if that is fine with you.
Sure. Thank you.
The build passed, merging this
Linux 5.9 introduced the close_range syscall to effectively close multiple file descriptors at once. Coincidentally, the index of this syscall is equal to the one of pidfd_mem.
As multiple programs, e.g., systemd, attempt to use this syscall without checking against the kernel version, pidfd_mem breaks the boot process on many newer systems.
The patch proposes to move the syscall index far ahead to avoid potential future regressions until kvmi is merged upstream.