I see I misunderstood, and this does what is desired.
Miod Vallat <[email protected]> wrote:
> The code responsible for filling a page with repeated copies of the
> signal trampoline code assumes that PAGE_SIZE % sigfillsz == 0.
>
> While this is true on all currently supported OpenBSD platforms, this
> might not be the case in the future (and isn't the case on some
> no-longer official platforms).
>
> The following diff makes sure that we don't try to write more than
> PAGE_SIZE bytes in this page. Another possibility would be to assert
> that PAGE_SIZE % sigfillsz == 0 and only apply this diff once it becomes
> truly needed.
>
> Index: sys/kern/kern_exec.c
> ===================================================================
> RCS file: /OpenBSD/src/sys/kern/kern_exec.c,v
> retrieving revision 1.208
> diff -u -p -r1.208 kern_exec.c
> --- sys/kern/kern_exec.c 2 Aug 2019 02:17:35 -0000 1.208
> +++ sys/kern/kern_exec.c 25 Nov 2019 10:09:48 -0000
> @@ -832,7 +832,7 @@ exec_sigcode_map(struct process *pr, str
> if (e->e_sigobject == NULL) {
> extern int sigfillsiz;
> extern u_char sigfill[];
> - size_t off;
> + size_t off, left;
> vaddr_t va;
> int r;
>
> @@ -846,8 +846,12 @@ exec_sigcode_map(struct process *pr, str
> return (ENOMEM);
> }
>
> - for (off = 0; off < round_page(sz); off += sigfillsiz)
> - memcpy((caddr_t)va + off, sigfill, sigfillsiz);
> + for (off = 0, left = round_page(sz); left != 0;
> + off += sigfillsiz) {
> + size_t chunk = ulmin(left, sigfillsiz);
> + memcpy((caddr_t)va + off, sigfill, chunk);
> + left -= chunk;
> + }
> memcpy((caddr_t)va, e->e_sigcode, sz);
> uvm_unmap(kernel_map, va, va + round_page(sz));
> }
>
