x86: Fix overlap of i386 CPU_ENTRY_AREA with FIX_BTMAP
PATCH 1/1 x86: Fix overlap of i386 CPU_ENTRY_AREA with FIX_BTMAP
Since commit 92a0f81d8957 ("x86/cpu_entry_area: Move it out of the fixmap"), i386's CPU_ENTRY_AREA has been mapped to the memory area just below FIXADDR_START. But already immediately before FIXADDR_START is the FIX_BTMAP area, which means that early_ioremap can collide with the entry area.
It's especially bad on PAE where FIX_BTMAP_BEGIN gets aligned to exactly match CPU_ENTRY_AREA_BASE, so the first early_ioremap slot clobbers the IDT and causes interrupts during early boot to reset the system.
The overlap wasn't a problem before the CPU entry area was introduced, as the fixmap has classically been preceded by the pkmap or vmalloc areas, neither of which is used until early_ioremap is out of the picture.
Relocate CPU_ENTRY_AREA to below FIX_BTMAP, not just below the permanent fixmap area.
---
This fixes a triple fault during early boot of 4.15 with PAE on x86 when an Intel IOMMU is present. The ACPI table retrieval inside pci_iommu_alloc() called early_ioremap and unmapped the IDT, and so the exception inside test_wp_bit triple faulted.
2MiB of i386 kernel address space is wasted in the FIX_BTMAP area that isn't used after boot, but I couldn't see a better place to put CPU_ENTRY_AREA. We can't stick it just before the pkmap, since pkmap doesn't exist when highmem is disabled. We could potentially shift fixmap down from the top of the address space, but that's already variable due to reserve_top_address. The only other option would be to shuffle it in between lowmem and vmalloc, which looks messy due to their variable sizes.
arch/x86/include/asm/fixmap.h | 6 ++++-- arch/x86/include/asm/pgtable_32_types.h | 5 +++-- 2 files changed, 7 insertions(+), 4 deletions(-)
arch/x86/include/asm/fixmap.h +4 -2
@@ -137,8 +137,10 @@ enum fixed_addresses {extern void reserve_top_address (unsigned long reserve); -#define FIXADDR_SIZE (__end_of_permanent_fixed_addresses << PAGE_SHIFT) -#define FIXADDR_START (FIXADDR_TOP - FIXADDR_SIZE) +#define FIXADDR_SIZE (__end_of_permanent_fixed_addresses << PAGE_SHIFT) +#define FIXADDR_START (FIXADDR_TOP - FIXADDR_SIZE) +#define FIXADDR_ALL_SIZE (__end_of_fixed_addresses << PAGE_SHIFT) +#define FIXADDR_ALL_START (FIXADDR_TOP - FIXADDR_ALL_SIZE)extern int fixmaps_set;
arch/x86/include/asm/pgtable_32_types.h +3 -2
@@ -44,8 +44,9 @@ extern bool __vmalloc_start_set; /* set once high_memory is set */*/ #define CPU_ENTRY_AREA_PAGES (NR_CPUS * 40) -#define CPU_ENTRY_AREA_BASE \ - ((FIXADDR_START - PAGE_SIZE * (CPU_ENTRY_AREA_PAGES + 1)) & PMD_MASK) +#define CPU_ENTRY_AREA_BASE \ + ((FIXADDR_ALL_START - PAGE_SIZE * (CPU_ENTRY_AREA_PAGES + 1)) \ + & PMD_MASK)#define PKMAP_BASE \ ((CPU_ENTRY_AREA_BASE - PAGE_SIZE) & PMD_MASK)
Signature · 1 line
2.15.1
Quoted 10 lines · William Grant
Since commit
92a0f81d8957("x86/cpu_entry_area: Move it out of the fixmap"), i386'sCPU_ENTRY_AREAhas been mapped to the memory area just belowFIXADDR_START. But already immediately beforeFIXADDR_STARTis theFIX_BTMAParea, which means thatearly_ioremapcan collide with the entry area.It's especially bad on PAE where
FIX_BTMAP_BEGINgets aligned to exactly matchCPU_ENTRY_AREA_BASE, so the firstearly_ioremapslot clobbers the IDT and causes interrupts during early boot to reset the system.
Quoted 7 lines
The overlap wasn't a problem before the CPU entry area was introduced, as the fixmap has classically been preceded by the pkmap or vmalloc areas, neither of which is used until
early_ioremapis out of the picture.Relocate
CPU_ENTRY_AREAto belowFIX_BTMAP, not just below the permanent fixmap area.
Nice catch. I'm sure I stared at that more than once....
Thanks,
tglx
tip-bot for William Grant
merge bot[tip:x86/pti] x86/mm: Fix overlap of i386 CPU_ENTRY_AREA with FIX_BTMAP
In reply to William Grantx86/mm: Fix overlap of i386 CPU_ENTRY_AREA with FIX_BTMAP
Commit-ID: 55f49fcb879fbeebf2a8c1ac7c9e6d90df55f798 Gitweb: https://git.kernel.org/tip/55f49fcb879fbeebf2a8c1ac7c9e6d90df55f798 Author: William Grant <[email protected]> AuthorDate: Tue, 30 Jan 2018 22:22:55 +1100 Committer: Thomas Gleixner <[email protected]> CommitDate: Tue, 30 Jan 2018 15:30:35 +0100
x86/mm: Fix overlap of i386 CPU_ENTRY_AREA with FIX_BTMAP
Since commit 92a0f81d8957 ("x86/cpu_entry_area: Move it out of the fixmap"), i386's CPU_ENTRY_AREA has been mapped to the memory area just below FIXADDR_START. But already immediately before FIXADDR_START is the FIX_BTMAP area, which means that early_ioremap can collide with the entry area.
It's especially bad on PAE where FIX_BTMAP_BEGIN gets aligned to exactly match CPU_ENTRY_AREA_BASE, so the first early_ioremap slot clobbers the IDT and causes interrupts during early boot to reset the system.
The overlap wasn't a problem before the CPU entry area was introduced, as the fixmap has classically been preceded by the pkmap or vmalloc areas, neither of which is used until early_ioremap is out of the picture.
Relocate CPU_ENTRY_AREA to below FIX_BTMAP, not just below the permanent fixmap area.
--- arch/x86/include/asm/fixmap.h | 6 ++++-- arch/x86/include/asm/pgtable_32_types.h | 5 +++-- 2 files changed, 7 insertions(+), 4 deletions(-)
arch/x86/include/asm/fixmap.h +4 -2
@@ -137,8 +137,10 @@ enum fixed_addresses {extern void reserve_top_address (unsigned long reserve); -#define FIXADDR_SIZE (__end_of_permanent_fixed_addresses << PAGE_SHIFT) -#define FIXADDR_START (FIXADDR_TOP - FIXADDR_SIZE) +#define FIXADDR_SIZE (__end_of_permanent_fixed_addresses << PAGE_SHIFT) +#define FIXADDR_START (FIXADDR_TOP - FIXADDR_SIZE) +#define FIXADDR_TOT_SIZE (__end_of_fixed_addresses << PAGE_SHIFT) +#define FIXADDR_TOT_START (FIXADDR_TOP - FIXADDR_TOT_SIZE)extern int fixmaps_set;
arch/x86/include/asm/pgtable_32_types.h +3 -2
@@ -44,8 +44,9 @@ extern bool __vmalloc_start_set; /* set once high_memory is set */*/ #define CPU_ENTRY_AREA_PAGES (NR_CPUS * 40) -#define CPU_ENTRY_AREA_BASE \ - ((FIXADDR_START - PAGE_SIZE * (CPU_ENTRY_AREA_PAGES + 1)) & PMD_MASK) +#define CPU_ENTRY_AREA_BASE \ + ((FIXADDR_TOT_START - PAGE_SIZE * (CPU_ENTRY_AREA_PAGES + 1)) \ + & PMD_MASK)#define PKMAP_BASE \ ((CPU_ENTRY_AREA_BASE - PAGE_SIZE) & PMD_MASK)