powerpc: Better handle "error" type page faults on book3s64
There are a number of conditions in the DSISR that represent conditions for which there is no point looking for a VMA or trying to update a PTE. In fact, POWER9 adds a few with bad AMOs, bad "paste" instruction etc...
This improve our existing code to use symbolic constants for all DSISR bits, properly avoid hash_page() for the new bits, and handle both the old and new bits early in do_page_fault()
--
This obsoletes
[PATCH] powerpc: Hande page faults for bad paste and bad AMO
Note: We should further update 32-bit etc... to do something similar, or even consolidate the code.
According to architectures back to 1.07 and 32-bit 6xx PEM, those low DSISR bits are 0 on all other implementations, and we do mask out the SRR1 bits on instruction interrupts, so I think such consolidation, at least for BookS should be trivial but I will leave it as a separate patch.
--- arch/powerpc/include/asm/reg.h | 18 ++++++++++++++++++ arch/powerpc/kernel/exceptions-64s.S | 6 ++++-- arch/powerpc/mm/fault.c | 21 +++++++++++++++++++++ 3 files changed, 43 insertions(+), 2 deletions(-)
arch/powerpc/include/asm/reg.h +18 -0
@@ -272,16 +272,34 @@#define SPRN_DAR 0x013/* Data Address Register */ #define SPRN_DBCR 0x136/* e300 Data Breakpoint Control Reg */ #define SPRN_DSISR 0x012/* Data Storage Interrupt Status Register */ +#define DSISR_BAD_DIRECT_ST 0x80000000/* Direct store error (obsolete) */ #define DSISR_NOHPTE 0x40000000/* no translation found */ +#define DSISR_ATT_CONFLICT 0x20000000/* Attribute conflicts (P9) */ #define DSISR_PROTFAULT 0x08000000/* protection fault */ #define DSISR_BADACCESS 0x04000000/* bad access to CI or G */ #define DSISR_ISSTORE 0x02000000/* access was a store */ #define DSISR_DABRMATCH 0x00400000/* hit data breakpoint */ #define DSISR_NOSEGMENT 0x00200000/* SLB miss */ #define DSISR_KEYFAULT 0x00200000/* Key fault */ +#define DSISR_BAD_EXT_CTRL 0x00100000/* External control error (obsolete) */ #define DSISR_UNSUPP_MMU 0x00080000/* Unsupported MMU config */ #define DSISR_SET_RC 0x00040000/* Failed setting of R/C bits */ #define DSISR_PGDIRFAULT 0x00020000/* Fault on page directory */ +#define DSISR_BAD_COPYPASTE 0x00000008/* Copy/Paste on wrong mem type */ +#define DSISR_BAD_AMO 0x00000004/* Incorrect AMO opcode */ +#define DSISR_BAD_CI_LDST 0x00000002/* CI load/store with DR=1 or in G=0 HV space */ + +/* Don't bother going to hash_page for any of these */ +#define DSISR_DONT_HASH (DSISR_BAD_DIRECT_ST | \ + DSISR_ATT_CONFLICT | \ + DSISR_BADACCESS | \ + DSISR_DABRMATCH | \ + DSISR_BAD_EXT_CTRL | \ + DSISR_UNSUPP_MMU | \ + DSISR_BAD_COPYPASTE | \ + DSISR_BAD_AMO | \ + DSISR_BAD_CI_LDST) +#define SPRN_TBRL 0x10C/* Time Base Read Lower Register (user, R/O) */ #define SPRN_TBRU 0x10D/* Time Base Read Upper Register (user, R/O) */ #define SPRN_CIR 0x11B/* Chip Information Register (hyper, R/0) */
arch/powerpc/kernel/exceptions-64s.S +4 -2
@@ -1413,8 +1413,10 @@ USE_TEXT_SECTION() .balign IFETCH_ALIGN_BYTES do_hash_page: #ifdef CONFIG_PPC_STD_MMU_64 - andis. r0,r4,0xa450 /* weird error? */ - bne- handle_page_fault /* if not, try to insert a HPTE */ + lis r0,DSISR_DONT_HASH@h + ori r0,r0,DSISR_DONT_HASH@l + and. r0,r0,r4 + bne- handle_page_fault CURRENT_THREAD_INFO(r11, r1) lwz r0,TI_PREEMPT(r11) /* If we're in an "NMI" */ andis. r0,r0,NMI_MASK@h /* (i.e. an irq when soft-disabled) */
arch/powerpc/mm/fault.c +21 -0
@@ -337,6 +337,27 @@ int do_page_fault(struct pt_regs *regs, unsigned long address, }#endif /* CONFIG_PPC_ICSWX */ +#ifdef CONFIG_PPC_STD_MMU_64 +/* +* These faults indicate various errors that can't be recovered +* by updating the PTE, there is no point handling them any later +* than here. I also noticed in some case the DAR isn't being set +* propertly (DSISR_BAD_COPYPASTE for example). +*/ +if (error_code & (DSISR_ATT_CONFLICT | +DSISR_BADACCESS | +DSISR_UNSUPP_MMU | +DSISR_BAD_COPYPASTE | +DSISR_BAD_AMO | +DSISR_BAD_CI_LDST )) { +if (user_mode (regs)) +_exception (SIGBUS , regs,BUS_OBJERR , address); +else + rc =SIGBUS ; + goto bail; + } +#endif /* CONFIG_PPC_STD_MMU_64 */ +if (notify_page_fault (regs)) goto bail;
Aneesh Kumar K.V
Reviewed-byQuoted 9 lines · Benjamin Herrenschmidt
There are a number of conditions in the DSISR that represent conditions for which there is no point looking for a VMA or trying to update a PTE. In fact, POWER9 adds a few with bad AMOs, bad "paste" instruction etc...
This improve our existing code to use symbolic constants for all DSISR bits, properly avoid
hash_page()for the new bits, and handle both the old and new bits early indo_page_fault()
Quoted 108 lines
Signed-off-by Benjamin Herrenschmidt <[email protected]>--
This obsoletes
[PATCH] powerpc: Hande page faults for bad paste and bad AMONote: We should further update 32-bit etc... to do something similar, or even consolidate the code.
According to architectures back to 1.07 and 32-bit 6xx PEM, those low DSISR bits are 0 on all other implementations, and we do mask out the SRR1 bits on instruction interrupts, so I think such consolidation, at least for BookS should be trivial but I will leave it as a separate patch.
--- arch/powerpc/include/asm/reg.h | 18 ++++++++++++++++++ arch/powerpc/kernel/exceptions-64s.S | 6 ++++-- arch/powerpc/mm/fault.c | 21 +++++++++++++++++++++ 3 files changed, 43 insertions(+), 2 deletions(-)arch/powerpc/include/asm/reg.h +18
-0@@ -272,16 +272,34 @@#define SPRN_DAR 0x013/* Data Address Register */ #define SPRN_DBCR 0x136/* e300 Data Breakpoint Control Reg */ #define SPRN_DSISR 0x012/* Data Storage Interrupt Status Register */ +#define DSISR_BAD_DIRECT_ST 0x80000000/* Direct store error (obsolete) */ #define DSISR_NOHPTE 0x40000000/* no translation found */ +#define DSISR_ATT_CONFLICT 0x20000000/* Attribute conflicts (P9) */ #define DSISR_PROTFAULT 0x08000000/* protection fault */ #define DSISR_BADACCESS 0x04000000/* bad access to CI or G */ #define DSISR_ISSTORE 0x02000000/* access was a store */ #define DSISR_DABRMATCH 0x00400000/* hit data breakpoint */ #define DSISR_NOSEGMENT 0x00200000/* SLB miss */ #define DSISR_KEYFAULT 0x00200000/* Key fault */ +#define DSISR_BAD_EXT_CTRL 0x00100000/* External control error (obsolete) */ #define DSISR_UNSUPP_MMU 0x00080000/* Unsupported MMU config */ #define DSISR_SET_RC 0x00040000/* Failed setting of R/C bits */ #define DSISR_PGDIRFAULT 0x00020000/* Fault on page directory */ +#define DSISR_BAD_COPYPASTE 0x00000008/* Copy/Paste on wrong mem type */ +#define DSISR_BAD_AMO 0x00000004/* Incorrect AMO opcode */ +#define DSISR_BAD_CI_LDST 0x00000002/* CI load/store with DR=1 or in G=0 HV space */ + +/* Don't bother going to hash_page for any of these */ +#define DSISR_DONT_HASH (DSISR_BAD_DIRECT_ST | \ + DSISR_ATT_CONFLICT | \ + DSISR_BADACCESS | \ + DSISR_DABRMATCH | \ + DSISR_BAD_EXT_CTRL | \ + DSISR_UNSUPP_MMU | \ + DSISR_BAD_COPYPASTE | \ + DSISR_BAD_AMO | \ + DSISR_BAD_CI_LDST) +#define SPRN_TBRL 0x10C/* Time Base Read Lower Register (user, R/O) */ #define SPRN_TBRU 0x10D/* Time Base Read Upper Register (user, R/O) */ #define SPRN_CIR 0x11B/* Chip Information Register (hyper, R/0) */ arch/powerpc/kernel/exceptions-64s.S +4
-2@@ -1413,8 +1413,10 @@ USE_TEXT_SECTION() .balign IFETCH_ALIGN_BYTES do_hash_page: #ifdef CONFIG_PPC_STD_MMU_64 - andis. r0,r4,0xa450 /* weird error? */ - bne- handle_page_fault /* if not, try to insert a HPTE */ + lis r0,DSISR_DONT_HASH@h + ori r0,r0,DSISR_DONT_HASH@l + and. r0,r0,r4 + bne- handle_page_fault CURRENT_THREAD_INFO(r11, r1) lwz r0,TI_PREEMPT(r11) /* If we're in an "NMI" */ andis. r0,r0,NMI_MASK@h /* (i.e. an irq when soft-disabled) */arch/powerpc/mm/fault.c +21
-0@@ -337,6 +337,27 @@ int do_page_fault(struct pt_regs *regs, unsigned long address, }#endif /* CONFIG_PPC_ICSWX */ +#ifdef CONFIG_PPC_STD_MMU_64 +/* +* These faults indicate various errors that can't be recovered +* by updating the PTE, there is no point handling them any later +* than here. I also noticed in some case the DAR isn't being set +* propertly (DSISR_BAD_COPYPASTE for example). +*/ +if (error_code & (DSISR_ATT_CONFLICT | +DSISR_BADACCESS | +DSISR_UNSUPP_MMU | +DSISR_BAD_COPYPASTE | +DSISR_BAD_AMO | +DSISR_BAD_CI_LDST )) { +if (user_mode (regs)) +_exception (SIGBUS , regs,BUS_OBJERR , address); +else + rc =SIGBUS ; + goto bail; + } +#endif /* CONFIG_PPC_STD_MMU_64 */ +if (notify_page_fault (regs)) goto bail;
Balbir Singh
Acked-byQuoted 5 lines · Benjamin Herrenschmidt
There are a number of conditions in the DSISR that represent conditions for which there is no point looking for a VMA or trying to update a PTE. In fact, POWER9 adds a few with bad AMOs, bad "paste" instruction etc...
Quoted 7 lines
This improve our existing code to use symbolic constants for all DSISR bits, properly avoid
hash_page()for the new bits, and handle both the old and new bits early indo_page_fault()Signed-off-by Benjamin Herrenschmidt <[email protected]>--
Looks reasonable