From: Heiko Carstens <hca@linux.ibm.com>
To: Alexander Gordeev <agordeev@linux.ibm.com>,
Sven Schnelle <svens@linux.ibm.com>,
Vasily Gorbik <gor@linux.ibm.com>,
Christian Borntraeger <borntraeger@linux.ibm.com>
Cc: linux-s390@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: [PATCH] s390/mm: Simplify crst_table_upgrade()
Date: Thu, 27 Aug 2026 12:28:32 +0200 [thread overview]
Message-ID: <20260827102832.3823857-1-hca@linux.ibm.com> (raw)
In case of an upgrade from four to five level page tables, and a failing
pgd allocation, the exit path of crst_table_upgrade() would incorrectly
dereference the p4d NULL pointer via pagetable_dtor().
Address this by reworking crst_table_upgrade(), which basically is a
revert of [1]. Take into account that GFP_KERNEL order-2 allocation
failures are very unlikely. Therefore keep the code as simple as
possible:
In case of an upgrade from three to five levels, and an allocation
failure of the fifth page table level, keep the upgrade to four levels
instead of reverting back to three levels. This allows to keep error
handling minimal.
[1] commit 31932757c612 ("s390/mm: optimize page table upgrade routine")
Reviewed-by: Alexander Gordeev <agordeev@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
---
arch/s390/mm/pgalloc.c | 89 +++++++++++++++++-------------------------
1 file changed, 36 insertions(+), 53 deletions(-)
diff --git a/arch/s390/mm/pgalloc.c b/arch/s390/mm/pgalloc.c
index 9610770fcf6d..4b160eedc5a0 100644
--- a/arch/s390/mm/pgalloc.c
+++ b/arch/s390/mm/pgalloc.c
@@ -55,63 +55,46 @@ static void __crst_table_upgrade(void *arg)
int crst_table_upgrade(struct mm_struct *mm, unsigned long end)
{
- unsigned long *pgd = NULL, *p4d = NULL, *__pgd;
- unsigned long asce_limit = mm->context.asce_limit;
+ unsigned long *table, *pgd;
+ int rc, notify;
mmap_assert_write_locked(mm);
-
/* upgrade should only happen from 3 to 4, 3 to 5, or 4 to 5 levels */
- VM_BUG_ON(asce_limit < _REGION2_SIZE);
-
- if (end <= asce_limit)
- return 0;
-
- if (asce_limit == _REGION2_SIZE) {
- p4d = crst_table_alloc(mm);
- if (unlikely(!p4d))
- goto err_p4d;
- crst_table_init(p4d, _REGION2_ENTRY_EMPTY);
- pagetable_p4d_ctor(virt_to_ptdesc(p4d));
+ VM_BUG_ON(mm->context.asce_limit < _REGION2_SIZE);
+ rc = 0;
+ notify = 0;
+ while (mm->context.asce_limit < end) {
+ table = crst_table_alloc(mm);
+ if (!table) {
+ rc = -ENOMEM;
+ break;
+ }
+ spin_lock_bh(&mm->page_table_lock);
+ pgd = (unsigned long *)mm->pgd;
+ if (mm->context.asce_limit == _REGION2_SIZE) {
+ crst_table_init(table, _REGION2_ENTRY_EMPTY);
+ p4d_populate(mm, (p4d_t *)table, (pud_t *)pgd);
+ pagetable_p4d_ctor(virt_to_ptdesc(table));
+ mm->pgd = (pgd_t *)table;
+ mm->context.asce_limit = _REGION1_SIZE;
+ mm->context.asce = __pa(mm->pgd) | _ASCE_TABLE_LENGTH |
+ _ASCE_USER_BITS | _ASCE_TYPE_REGION2;
+ mm_inc_nr_puds(mm);
+ } else {
+ crst_table_init(table, _REGION1_ENTRY_EMPTY);
+ pgd_populate(mm, (pgd_t *)table, (p4d_t *)pgd);
+ pagetable_pgd_ctor(virt_to_ptdesc(table));
+ mm->pgd = (pgd_t *)table;
+ mm->context.asce_limit = TASK_SIZE_MAX;
+ mm->context.asce = __pa(mm->pgd) | _ASCE_TABLE_LENGTH |
+ _ASCE_USER_BITS | _ASCE_TYPE_REGION1;
+ }
+ notify = 1;
+ spin_unlock_bh(&mm->page_table_lock);
}
- if (end > _REGION1_SIZE) {
- pgd = crst_table_alloc(mm);
- if (unlikely(!pgd))
- goto err_pgd;
- crst_table_init(pgd, _REGION1_ENTRY_EMPTY);
- pagetable_pgd_ctor(virt_to_ptdesc(pgd));
- }
-
- spin_lock_bh(&mm->page_table_lock);
-
- if (p4d) {
- __pgd = (unsigned long *) mm->pgd;
- p4d_populate(mm, (p4d_t *) p4d, (pud_t *) __pgd);
- mm->pgd = (pgd_t *) p4d;
- mm->context.asce_limit = _REGION1_SIZE;
- mm->context.asce = __pa(mm->pgd) | _ASCE_TABLE_LENGTH |
- _ASCE_USER_BITS | _ASCE_TYPE_REGION2;
- mm_inc_nr_puds(mm);
- }
- if (pgd) {
- __pgd = (unsigned long *) mm->pgd;
- pgd_populate(mm, (pgd_t *) pgd, (p4d_t *) __pgd);
- mm->pgd = (pgd_t *) pgd;
- mm->context.asce_limit = TASK_SIZE_MAX;
- mm->context.asce = __pa(mm->pgd) | _ASCE_TABLE_LENGTH |
- _ASCE_USER_BITS | _ASCE_TYPE_REGION1;
- }
-
- spin_unlock_bh(&mm->page_table_lock);
-
- on_each_cpu(__crst_table_upgrade, mm, 0);
-
- return 0;
-
-err_pgd:
- pagetable_dtor(virt_to_ptdesc(p4d));
- crst_table_free(mm, p4d);
-err_p4d:
- return -ENOMEM;
+ if (notify)
+ on_each_cpu(__crst_table_upgrade, mm, 0);
+ return rc;
}
unsigned long *page_table_alloc_noprof(struct mm_struct *mm)
--
2.53.0
next reply other threads:[~2026-08-27 10:28 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-08-27 10:28 Heiko Carstens [this message]
2026-08-27 10:37 ` [PATCH] s390/mm: Simplify crst_table_upgrade() sashiko-bot
2026-08-27 15:59 ` Vasily Gorbik
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20260827102832.3823857-1-hca@linux.ibm.com \
--to=hca@linux.ibm.com \
--cc=agordeev@linux.ibm.com \
--cc=borntraeger@linux.ibm.com \
--cc=gor@linux.ibm.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-s390@vger.kernel.org \
--cc=svens@linux.ibm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox