On Tue, Jul 18, 2023 at 05:51:54PM +0000, Edgecombe, Rick P wrote:
On Sun, 2023-07-16 at 22:51 +0100, Mark Brown wrote:
+ gcs = alloc_gcs(task->thread.gcspr_el0, size, + 0, 0); + if (!gcs) + return -ENOMEM;
+ task->thread.gcs_el0_mode = arg; + if (task == current) + gcs_set_el0_mode(task);
Ah! So does this task == current part mean this can be called from another task via ptrace?
Ugh, right. So I had been worried about preemption rather than invoking prctl() via ptrace, though since ptrace can change the syscall invoked by a task it could cause prctl() to be invoked that way (but that should look like being run in the target process).
I'm not aware of an interface specifically intended to remotely invoke prctls via ptrace but that doesn't mean there isn't one that I didn't find yet. I can't remember why I'm aware of the task != current case as a concern which is worrying me.
If so, then is the alloc_gcs() part on the wrong mm?
Yes, it will be. I'll add a check in there to reject attempts to enable GCS when task != current.