Re: git: 21e45c30c35c - main - mmap(MAP_STACK): on stack grow, use original protection
Date: Thu, 20 Jul 2023 16:32:31 UTC
In message <202307201412.36KECDSU084918@gitrepo.freebsd.org>, Konstantin
Belous
ov writes:
> The branch main has been updated by kib:
>
> URL: https://cgit.FreeBSD.org/src/commit/?id=21e45c30c35c9aa732073f725924caf5
> 81c93460
>
> commit 21e45c30c35c9aa732073f725924caf581c93460
> Author: Konstantin Belousov <kib@FreeBSD.org>
> AuthorDate: 2023-07-19 11:05:32 +0000
> Commit: Konstantin Belousov <kib@FreeBSD.org>
> CommitDate: 2023-07-20 14:11:42 +0000
>
> mmap(MAP_STACK): on stack grow, use original protection
>
> If mprotect(2) changed protection in the bottom of the currently grown
> stack region, currently the changed protection would be used for the
> stack grow on next fault. This is arguably unexpected.
>
> Store the original protection for the entry at mmap(2) time in the
> offset member of the gap vm_map_entry, and use it for protection of the
> grown stack region.
>
> PR: 272585
> Reported by: John F. Carr <jfc@mit.edu>
> Reviewed by: alc, markj
> Sponsored by: The FreeBSD Foundation
> MFC after: 1 week
> Differential revision: https://reviews.freebsd.org/D41089
> ---
> sys/vm/vm_map.c | 24 ++++++++++++++++--------
> sys/vm/vm_map.h | 4 ++++
> 2 files changed, 20 insertions(+), 8 deletions(-)
>
> diff --git a/sys/vm/vm_map.c b/sys/vm/vm_map.c
> index a02107b5e64d..997a49111a59 100644
> --- a/sys/vm/vm_map.c
> +++ b/sys/vm/vm_map.c
> @@ -4493,7 +4493,7 @@ static int
> vm_map_stack_locked(vm_map_t map, vm_offset_t addrbos, vm_size_t max_ssize,
> vm_size_t growsize, vm_prot_t prot, vm_prot_t max, int cow)
> {
> - vm_map_entry_t new_entry, prev_entry;
> + vm_map_entry_t gap_entry, new_entry, prev_entry;
> vm_offset_t bot, gap_bot, gap_top, top;
> vm_size_t init_ssize, sgp;
> int orient, rv;
> @@ -4575,11 +4575,14 @@ vm_map_stack_locked(vm_map_t map, vm_offset_t addrbos
> , vm_size_t max_ssize,
> * read-ahead logic is never used for it. Re-use
> * next_read of the gap entry to store
> * stack_guard_page for vm_map_growstack().
> + * Similarly, since a gap cannot have a backing object,
> + * store the original stack protections in the
> + * object offset.
> */
> - if (orient == MAP_STACK_GROWS_DOWN)
> - vm_map_entry_pred(new_entry)->next_read = sgp;
> - else
> - vm_map_entry_succ(new_entry)->next_read = sgp;
> + gap_entry = orient == MAP_STACK_GROWS_DOWN ?
> + vm_map_entry_pred(new_entry) : vm_map_entry_succ(new_entry)
> ;
> + gap_entry->next_read = sgp;
> + gap_entry->offset = prot;
> } else {
> (void)vm_map_delete(map, bot, top);
> }
> @@ -4599,6 +4602,7 @@ vm_map_growstack(vm_map_t map, vm_offset_t addr, vm_map
> _entry_t gap_entry)
> struct ucred *cred;
> vm_offset_t gap_end, gap_start, grow_start;
> vm_size_t grow_amount, guard, max_grow;
> + vm_prot_t prot;
> rlim_t lmemlim, stacklim, vmemlim;
> int rv, rv1 __diagused;
> bool gap_deleted, grow_down, is_procstack;
> @@ -4739,6 +4743,12 @@ retry:
> }
>
> if (grow_down) {
> + /*
> + * The gap_entry "offset" field is overloaded. See
> + * vm_map_stack_locked().
> + */
> + prot = gap_entry->offset;
> +
> grow_start = gap_entry->end - grow_amount;
> if (gap_entry->start + grow_amount == gap_entry->end) {
> gap_start = gap_entry->start;
> @@ -4751,9 +4761,7 @@ retry:
> gap_deleted = false;
> }
> rv = vm_map_insert(map, NULL, 0, grow_start,
> - grow_start + grow_amount,
> - stack_entry->protection, stack_entry->max_protection,
> - MAP_STACK_GROWS_DOWN);
> + grow_start + grow_amount, prot, prot, MAP_STACK_GROWS_DOWN)
> ;
> if (rv != KERN_SUCCESS) {
> if (gap_deleted) {
> rv1 = vm_map_insert(map, NULL, 0, gap_start,
> diff --git a/sys/vm/vm_map.h b/sys/vm/vm_map.h
> index fd8b606e8ddc..c4ed36ce57ba 100644
> --- a/sys/vm/vm_map.h
> +++ b/sys/vm/vm_map.h
> @@ -97,6 +97,10 @@ union vm_map_object {
> * a VM object (or sharing map) and offset into that object,
> * and user-exported inheritance and protection information.
> * Also included is control information for virtual copy operations.
> + *
> + * For stack gap map entries (MAP_ENTRY_GUARD | MAP_ENTRY_GROWS_DOWN
> + * or UP), the next_read member is reused as the stack_guard_page
> + * storage, and offset is the stack protection.
> */
> struct vm_map_entry {
> struct vm_map_entry *left; /* left child or previous entry */
>
This commit has caused screen (sysutils/screen) to display the following
error when opening a new screen window (ctrl-A c):
pid 28473 comm screen-4.9.0 has trashed its stack, killing
--
Cheers,
Cy Schubert <Cy.Schubert@cschubert.com>
FreeBSD UNIX: <cy@FreeBSD.org> Web: https://FreeBSD.org
NTP: <cy@nwtime.org> Web: https://nwtime.org
e^(i*pi)+1=0