Skip to contentSkip to conversation

x86/boot: Fix off-by-one error in sme_clear_pgd()

PATCH RESEND Applied Thorsten Blum 4 participants 6 messages Sep 18 – Nov 6 x86

lore
Share & tools
lore
#0 lore raw

PATCH RESEND 1/1 x86/boot: Fix off-by-one error in sme_clear_pgd()

The mapping helpers __sme_map_range_pmd() and __sme_map_range_pte(), and therefore sme_encrypt_kernel(), use exclusive end addresses and stop before vaddr reaches vaddr_end.

However, sme_clear_pgd() treats vaddr_end as inclusive when calculating the last PGD entry to clear. If vaddr_end is PGD-aligned, the first PGD entry outside the range is cleared too.

Use vaddr_end - 1 to calculate the last PGD entry within the range.

Fixes 6ebcb060713f ("x86/mm: Add support to encrypt the kernel in-place")
---
 arch/x86/boot/startup/sme.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/boot/startup/sme.c +1 -1
index c07a2c381ed1..2d5cb3f7eee6 100644
@@ -99,7 +99,7 @@ static void __init sme_clear_pgd(struct sme_populate_pgd_data *ppd)
 	pgd_t *pgd_p;
 
 	pgd_start = ppd->vaddr & PGDIR_MASK;
-	pgd_end = ppd->vaddr_end & PGDIR_MASK;
+	pgd_end = (ppd->vaddr_end - 1) & PGDIR_MASK;
 
 	pgd_size = (((pgd_end - pgd_start) / PGDIR_SIZE) + 1) * sizeof(pgd_t);
 
#1 lore raw
Quoted 9 lines · On Fri, 18 Sep 2026 16:27:13 +0200 Thorsten Blum <[email protected]> wrote:

The mapping helpers __sme_map_range_pmd() and __sme_map_range_pte(), and therefore sme_encrypt_kernel(), use exclusive end addresses and stop before vaddr reaches vaddr_end.

However, sme_clear_pgd() treats vaddr_end as inclusive when calculating the last PGD entry to clear. If vaddr_end is PGD-aligned, the first PGD entry outside the range is cleared too.

Use vaddr_end - 1 to calculate the last PGD entry within the range.

When fixing a bug, please always fully describe the worst-case userspace-visible runtime effects of that bug.

I'm guessing "none known, this is from code inspection"?

Thanks.

Quoted 11 lines
arch/x86/boot/startup/sme.c +1 -1
@@ -99,7 +99,7 @@ static void __init sme_clear_pgd(struct sme_populate_pgd_data *ppd)
 	pgd_t *pgd_p;
 
 	pgd_start = ppd->vaddr & PGDIR_MASK;
-	pgd_end = ppd->vaddr_end & PGDIR_MASK;
+	pgd_end = (ppd->vaddr_end - 1) & PGDIR_MASK;
 
 	pgd_size = (((pgd_end - pgd_start) / PGDIR_SIZE) + 1) * sizeof(pgd_t);
 
#2 lore raw
Quoted 10 lines · Andrew Morton
Quoted 8 lines · On Fri, 18 Sep 2026 16:27:13 +0200 Thorsten Blum <[email protected]> wrote:

The mapping helpers __sme_map_range_pmd() and __sme_map_range_pte(), and therefore sme_encrypt_kernel(), use exclusive end addresses and stop before vaddr reaches vaddr_end.

However, sme_clear_pgd() treats vaddr_end as inclusive when calculating the last PGD entry to clear. If vaddr_end is PGD-aligned, the first PGD entry outside the range is cleared too.

Quoted 6 lines
Quoted 1 line

Use vaddr_end - 1 to calculate the last PGD entry within the range.

When fixing a bug, please always fully describe the worst-case userspace-visible runtime effects of that bug.

I'm guessing "none known, this is from code inspection"?

Correct, this was found by code inspection.

Reviewed-by on v1 1/1
#3 lore raw
In reply to Thorsten Blum
Quoted 4 lines · Thorsten Blum

The mapping helpers __sme_map_range_pmd() and __sme_map_range_pte(), and therefore sme_encrypt_kernel(), use exclusive end addresses and stop before vaddr reaches vaddr_end.

Quoted 8 lines

However, sme_clear_pgd() treats vaddr_end as inclusive when calculating the last PGD entry to clear. If vaddr_end is PGD-aligned, the first PGD entry outside the range is cleared too.

Use vaddr_end - 1 to calculate the last PGD entry within the range.

Fixes 6ebcb060713f ("x86/mm: Add support to encrypt the kernel in-place")
Quoted 17 lines
---
 arch/x86/boot/startup/sme.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/boot/startup/sme.c +1 -1
index c07a2c381ed1..2d5cb3f7eee6 100644
@@ -99,7 +99,7 @@ static void __init sme_clear_pgd(struct sme_populate_pgd_data *ppd)
 	pgd_t *pgd_p;
 
 	pgd_start = ppd->vaddr & PGDIR_MASK;
-	pgd_end = ppd->vaddr_end & PGDIR_MASK;
+	pgd_end = (ppd->vaddr_end - 1) & PGDIR_MASK;
 
 	pgd_size = (((pgd_end - pgd_start) / PGDIR_SIZE) + 1) * sizeof(pgd_t);
 
merge bot Applied
#4 lore raw
In reply to Thorsten Blum
The following commit has been merged into the x86/boot branch of tip:

The following commit has been merged into the x86/boot branch of tip:

Commit-ID:     8523d938dce836fbe7ea33fcc98d8460b094ef81
Gitweb:        https://git.kernel.org/tip/8523d938dce836fbe7ea33fcc98d8460b094ef81
Author:        Thorsten Blum <[email protected]>
AuthorDate:    Fri, 18 Sep 2026 16:27:13 +02:00
Committer:     Borislav Petkov (AMD) <[email protected]>
CommitterDate: Thu, 05 Nov 2026 10:21:38 -08:00

x86/boot: Fix off-by-one error in sme_clear_pgd()

The mapping helpers __sme_map_range_pmd() and __sme_map_range_pte(), and therefore sme_encrypt_kernel(), use exclusive end addresses and stop before vaddr reaches vaddr_end.

However, sme_clear_pgd() treats vaddr_end as inclusive when calculating the last PGD entry to clear. If vaddr_end is PGD-aligned, the first PGD entry outside the range is cleared too.

Use vaddr_end - 1 to calculate the last PGD entry within the range.

Fixes 6ebcb060713f ("x86/mm: Add support to encrypt the kernel in-place")
---
 arch/x86/boot/startup/sme.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/boot/startup/sme.c +1 -1
index c07a2c3..2d5cb3f 100644
@@ -99,7 +99,7 @@ static void __init sme_clear_pgd(struct sme_populate_pgd_data *ppd)
 	pgd_t *pgd_p;
 
 	pgd_start = ppd->vaddr & PGDIR_MASK;
-	pgd_end = ppd->vaddr_end & PGDIR_MASK;
+	pgd_end = (ppd->vaddr_end - 1) & PGDIR_MASK;
 
 	pgd_size = (((pgd_end - pgd_start) / PGDIR_SIZE) + 1) * sizeof(pgd_t);
 
merge bot Applied
#5 lore raw
In reply to Thorsten Blum
The following commit has been merged into the x86/urgent branch of tip:

The following commit has been merged into the x86/urgent branch of tip:

Commit-ID:     cb319926808b1cca522d2087a13205e840beb45b
Gitweb:        https://git.kernel.org/tip/cb319926808b1cca522d2087a13205e840beb45b
Author:        Thorsten Blum <[email protected]>
AuthorDate:    Fri, 18 Sep 2026 16:27:13 +02:00
Committer:     Borislav Petkov (AMD) <[email protected]>
CommitterDate: Fri, 06 Nov 2026 09:05:23 -08:00

x86/boot: Fix off-by-one error in sme_clear_pgd()

The mapping helpers __sme_map_range_pmd() and __sme_map_range_pte(), and therefore sme_encrypt_kernel(), use exclusive end addresses and stop before vaddr reaches vaddr_end.

However, sme_clear_pgd() treats vaddr_end as inclusive when calculating the last PGD entry to clear. If vaddr_end is PGD-aligned, the first PGD entry outside the range is cleared too.

Use vaddr_end - 1 to calculate the last PGD entry within the range.

Fixes 6ebcb060713f ("x86/mm: Add support to encrypt the kernel in-place")
---
 arch/x86/boot/startup/sme.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
arch/x86/boot/startup/sme.c +1 -1
index c07a2c3..2d5cb3f 100644
@@ -99,7 +99,7 @@ static void __init sme_clear_pgd(struct sme_populate_pgd_data *ppd)
 	pgd_t *pgd_p;
 
 	pgd_start = ppd->vaddr & PGDIR_MASK;
-	pgd_end = ppd->vaddr_end & PGDIR_MASK;
+	pgd_end = (ppd->vaddr_end - 1) & PGDIR_MASK;
 
 	pgd_size = (((pgd_end - pgd_start) / PGDIR_SIZE) + 1) * sizeof(pgd_t);