The 08/09/2023 16:34, Mark Brown wrote: > On Wed, Aug 09, 2023 at 03:24:14PM +0100, Catalin Marinas wrote: > > On Mon, Aug 07, 2023 at 11:00:08PM +0100, Mark Brown wrote: > > > +* When GCS is enabled for a thread a new Guarded Control Stack will be > > > + allocated for it of size RLIMIT_STACK / 2 or 2 gigabytes, whichever is > > > + smaller. > > > Is this number based on the fact that a function call would only push > > the LR to GCS while standard function prologue pushes at least two > > registers? > > It's actually based on bitrot that I'd initially chosen a smaller value > since it's likely that functions will push at least something as you > suggest, the patches now just use RLIMIT_STACK. I'll fix. the pcs requires 16byte aligned stack frames, with 8byte per gcs entry there is no need for same gcs size as stack size in userspace. you can argue about a fixed size small increment (stacksize/2 + inc) for signal handling on alt stack and special tokens, but stack size is overkill i think. fwiw my current makecontext patch uses roundup(stacksize/2+160). (threads guaranteed to have about 300bytes of data on the stack in glibc so if gcs is stacksize/2, that accounts for the increment. this is for the theoretical case when an empty thread just tries to overflow the stack and then handle the fault on sigaltstack.)
