series to optionally disable tlbie for 64s/radix
Subsystems (5)
Nicholas Piggin
reviewercover series to optionally disable tlbie for 64s/radix
Since the RFC I accounted for feedback, and also made the patch more complete so we can actually boot and run a 64s/radix kernel without using tlbie at all.
KVM and bare metal hash are harder to support. Bare metal hash because it does some TLB invation with interrupts disabled (can't use IPIs). It might be possible those invalidates could be avoided or the paths changed to enable interrupts, but it would be a much bigger change.
KVM with radix guests might be acutally quite simple to support, so I can look at that next if we get this merged.
KVM with hash guests might not be feasible to do in a performant way, because of the design of virtualised hash architecture (host flushes guest effective addresses), at least it would need special real mode IPI handlers for TLB flushes.
Thanks,
Nick
Nicholas Piggin (3):
powerpc/64s: Remove mmu_partition_table_set powerpc/64s/radix: all CPUs should flush local translation structure
before turning MMU on
powerpc/64s: introduce options to disable use of the tlbie instruction
.../admin-guide/kernel-parameters.txt | 4 + arch/powerpc/include/asm/book3s/64/tlbflush.h | 9 + arch/powerpc/include/asm/mmu.h | 2 - arch/powerpc/kvm/book3s_hv.c | 10 +- arch/powerpc/kvm/book3s_hv_nested.c | 35 +++- arch/powerpc/mm/book3s64/hash_utils.c | 4 +- arch/powerpc/mm/book3s64/mmu_context.c | 4 +- arch/powerpc/mm/book3s64/pgtable.c | 77 +++++--- arch/powerpc/mm/book3s64/radix_pgtable.c | 22 +-- arch/powerpc/mm/book3s64/radix_tlb.c | 182 ++++++++++++++++-- arch/powerpc/mm/pgtable_64.c | 5 +- drivers/misc/cxl/main.c | 4 + drivers/misc/ocxl/main.c | 4 + 13 files changed, 282 insertions(+), 80 deletions(-)
Signature · 1 line
2.22.0
Nicholas Piggin
reviewerPATCH 1/3 powerpc/64s: Remove mmu_partition_table_set
This was an awkward interface that didn't quite suit bare metal boot, because the host page table base is stored into the partition table at a different place than the process table base, and the TLB flushing is unnecessary at boot with the MMU off, provided all CPUs invalidate all their translation structures before turning their MMU on.
So inline this into kvmhv_set_ptbl_entry, where it pairs up nicely with the nested case.
Bare metal modifies the partition table directly, and the TLBIEs are dropped because in HVMODE, all CPUs invalidate their translation structures before enabling their MMU.
--- arch/powerpc/include/asm/mmu.h | 2 -- arch/powerpc/kvm/book3s_hv.c | 4 +-- arch/powerpc/kvm/book3s_hv_nested.c | 35 +++++++++++++++++++----- arch/powerpc/mm/book3s64/hash_utils.c | 4 ++- arch/powerpc/mm/book3s64/mmu_context.c | 4 +-- arch/powerpc/mm/book3s64/pgtable.c | 30 -------------------- arch/powerpc/mm/book3s64/radix_pgtable.c | 11 +++----- arch/powerpc/mm/pgtable_64.c | 5 ++-- 8 files changed, 42 insertions(+), 53 deletions(-)
arch/powerpc/include/asm/mmu.h +0 -2
@@ -256,8 +256,6 @@ extern void radix__mmu_cleanup_all(void);/* Functions for creating and updating partition table on POWER9 */ extern void mmu_partition_table_init (void ); -extern void mmu_partition_table_set_entry (unsigned int lpid,unsigned long dw0, -unsigned long dw1);#endif /* CONFIG_PPC64 */ struct mm_struct ;
arch/powerpc/kvm/book3s_hv.c +2 -2
@@ -4786,8 +4786,8 @@ static int kvmppc_core_init_vm_hv(struct kvm *kvm)* Since we don't flush the TLB when tearing down a VM, * and this lpid might have previously been used, * make sure we flush on each core before running the new VM. -* On POWER9, the tlbie in mmu_partition_table_set_entry() -* does this flush for us. +* On POWER9, the flush setting the partition table entry +* in kvmhv_set_ptbl_entry takes care of this. */ if (!cpu_has_feature (CPU_FTR_ARCH_300 ))cpumask_setall (&kvm->arch.need_tlb_flush);
arch/powerpc/kvm/book3s_hv_nested.c +28 -7
@@ -411,14 +411,35 @@ static void kvmhv_flush_lpid(unsigned int lpid)void kvmhv_set_ptbl_entry (unsigned int lpid,u64 dw0,u64 dw1) {if (!kvmhv_on_pseries ()) { -mmu_partition_table_set_entry (lpid, dw0, dw1); -return ; - } +unsigned long old =be64_to_cpu (partition_tb[lpid].patb0); + + partition_tb[lpid].patb0 =cpu_to_be64 (dw0); + partition_tb[lpid].patb1 =cpu_to_be64 (dw1); + +/* +* Global flush of TLBs and partition table caches for this +* lpid. The type of flush (hash or radix) depends on what the +* previous use of this partition ID was, not the new use. +*/ + asmvolatile ("ptesync" : : :"memory" ); +if (old &PATB_HR ) { + asm volatile(PPC_TLBIE_5(%0,%1,2,0,1) : : + "r" (TLBIEL_INVAL_SET_LPID), "r" (lpid)); + asm volatile(PPC_TLBIE_5(%0,%1,2,1,1) : : + "r" (TLBIEL_INVAL_SET_LPID), "r" (lpid)); + }else { + asm volatile(PPC_TLBIE_5(%0,%1,2,0,0) : : + "r" (TLBIEL_INVAL_SET_LPID), "r" (lpid)); + } +/* do we need fixup here ?*/ + asmvolatile ("eieio; tlbsync; ptesync" : : :"memory" ); - pseries_partition_tb[lpid].patb0 =cpu_to_be64 (dw0); - pseries_partition_tb[lpid].patb1 =cpu_to_be64 (dw1); -/* L0 will do the necessary barriers */ -kvmhv_flush_lpid (lpid); + }else { + pseries_partition_tb[lpid].patb0 =cpu_to_be64 (dw0); + pseries_partition_tb[lpid].patb1 =cpu_to_be64 (dw1); +/* L0 will do the necessary barriers */ +kvmhv_flush_lpid (lpid); + } }static void kvmhv_set_nested_ptbl (struct kvm_nested_guest *gp)
arch/powerpc/mm/book3s64/hash_utils.c +3 -1
@@ -823,7 +823,9 @@ static void __init hash_init_partition_table(phys_addr_t hash_table,* For now, UPRT is 0 and we have no segment table. */ htab_size =__ilog2 (htab_size) -18 ; -mmu_partition_table_set_entry (0 , hash_table | htab_size,0 ); + partition_tb[0 ].patb0 =cpu_to_be64 (hash_table | htab_size); + partition_tb[0 ].patb1 =0 ; +pr_info ("Partition table %p\n" , partition_tb); }
arch/powerpc/mm/book3s64/mmu_context.c +2 -2
@@ -276,8 +276,8 @@ void arch_exit_mmap(struct mm_struct *mm)* and 0 is invalid. So this will do. * * This runs before the "fullmm" tlb flush in exit_mmap, -* which does a RIC=2 tlbie to clear the process table -* entry. See the "fullmm" comments in tlb-radix.c. +* which does a RIC=2 tlb invalidate to clear the process +* table entry. See the "fullmm" comments in tlb-radix.c. * * No barrier required here after the store because * this process will do the invalidate, which starts with
arch/powerpc/mm/book3s64/pgtable.c +0 -30
@@ -209,36 +209,6 @@ void __init mmu_partition_table_init(void)powernv_set_nmmu_ptcr (ptcr); } -void mmu_partition_table_set_entry (unsigned int lpid,unsigned long dw0, -unsigned long dw1) -{ -unsigned long old =be64_to_cpu (partition_tb[lpid].patb0); - - partition_tb[lpid].patb0 =cpu_to_be64 (dw0); - partition_tb[lpid].patb1 =cpu_to_be64 (dw1); - -/* -* Global flush of TLBs and partition table caches for this lpid. -* The type of flush (hash or radix) depends on what the previous -* use of this partition ID was, not the new use. -*/ - asmvolatile ("ptesync" : : :"memory" ); -if (old &PATB_HR ) { - asm volatile(PPC_TLBIE_5(%0,%1,2,0,1) : : - "r" (TLBIEL_INVAL_SET_LPID), "r" (lpid)); - asm volatile(PPC_TLBIE_5(%0,%1,2,1,1) : : - "r" (TLBIEL_INVAL_SET_LPID), "r" (lpid)); -trace_tlbie (lpid,0 ,TLBIEL_INVAL_SET_LPID , lpid,2 ,0 ,1 ); - }else { - asm volatile(PPC_TLBIE_5(%0,%1,2,0,0) : : - "r" (TLBIEL_INVAL_SET_LPID), "r" (lpid)); -trace_tlbie (lpid,0 ,TLBIEL_INVAL_SET_LPID , lpid,2 ,0 ,0 ); - } -/* do we need fixup here ?*/ - asmvolatile ("eieio; tlbsync; ptesync" : : :"memory" ); -} -EXPORT_SYMBOL_GPL (mmu_partition_table_set_entry); -static pmd_t *get_pmd_from_cache (struct mm_struct *mm) {void *pmd_frag, *ret;
arch/powerpc/mm/book3s64/radix_pgtable.c +4 -7
@@ -36,12 +36,7 @@ unsigned int mmu_base_pid;static int native_register_process_table (unsigned long base,unsigned long pg_sz,unsigned long table_size) { -unsigned long patb0, patb1; - - patb0 =be64_to_cpu (partition_tb[0 ].patb0); - patb1 = base | table_size |PATB_GR ; - -mmu_partition_table_set_entry (0 , patb0, patb1); + partition_tb[0 ].patb1 =cpu_to_be64 (base | table_size |PATB_GR );return 0 ; }
@@ -417,7 +412,9 @@ static void __init radix_init_partition_table(void)mmu_partition_table_init (); rts_field =radix__get_tree_size (); dw0 = rts_field |__pa (init_mm.pgd) |RADIX_PGD_INDEX_SIZE |PATB_HR ; -mmu_partition_table_set_entry (0 , dw0,0 ); + + partition_tb[0 ].patb0 =cpu_to_be64 (dw0); + partition_tb[0 ].patb1 =0 ;pr_info ("Initializing Radix MMU\n" );pr_info ("Partition table %p\n" , partition_tb);
arch/powerpc/mm/pgtable_64.c +3 -2
@@ -54,8 +54,9 @@/* * partition table and process table for ISA 3.0 */ -struct prtb_entry *process_tb; -struct patb_entry *partition_tb; +struct prtb_entry *process_tb__read_mostly ; +struct patb_entry *partition_tb__read_mostly ; +EXPORT_SYMBOL_GPL (partition_tb);/* KVM only */ /* * page table size */
Signature · 1 line
2.22.0
Nicholas Piggin
reviewerPATCH 2/3 powerpc/64s/radix: all CPUs should flush local translation structure before turning MMU
In reply to Nicholas PigginRather than sprinkle various translation structure invalidations around different places in early boot, have each CPU flush everything from its local translation structures before enabling its MMU.
Radix guests can execute tlbie(l), so have them tlbiel_all in the same place as radix host does.
--- arch/powerpc/mm/book3s64/radix_pgtable.c | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-)
arch/powerpc/mm/book3s64/radix_pgtable.c +2 -9
@@ -382,11 +382,6 @@ static void __init radix_init_pgtable(void)*/ register_process_table (__pa (process_tb),0 ,PRTB_SIZE_SHIFT -12 );pr_info ("Process table %p and radix root for kernel: %p\n" , process_tb, init_mm.pgd); - asmvolatile ("ptesync" : : :"memory" ); - asm volatile(PPC_TLBIE_5(%0,%1,2,1,1) : : - "r" (TLBIEL_INVAL_SET_LPID), "r" (0)); - asmvolatile ("eieio; tlbsync; ptesync" : : :"memory" ); -trace_tlbie (0 ,0 ,TLBIEL_INVAL_SET_LPID ,0 ,2 ,1 ,1 );/* * The init_mm context is given the first available (non-zero) PID,
@@ -633,8 +628,7 @@ void __init radix__early_init_mmu(void)radix_init_pgtable ();/* Switch to the guard PID before turning on MMU */ radix__switch_mmu_context (NULL , &init_mm); -if (cpu_has_feature (CPU_FTR_HVMODE )) -tlbiel_all (); +tlbiel_all (); }void radix__early_init_mmu_secondary (void )
@@ -653,8 +647,7 @@ void radix__early_init_mmu_secondary(void) }radix__switch_mmu_context (NULL , &init_mm); -if (cpu_has_feature (CPU_FTR_HVMODE )) -tlbiel_all (); +tlbiel_all (); }void radix__mmu_cleanup_all (void )
Signature · 1 line
2.22.0
Michael Ellerman
reviewerRe: [PATCH 2/3] powerpc/64s/radix: all CPUs should flush local translation structure before turning
Quoted 29 lines · Nicholas Piggin
Rather than sprinkle various translation structure invalidations around different places in early boot, have each CPU flush everything from its local translation structures before enabling its MMU.
Radix guests can execute tlbie(l), so have them
tlbiel_allin the same place as radix host does.Signed-off-by Nicholas Piggin <[email protected]>--- arch/powerpc/mm/book3s64/radix_pgtable.c | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-)arch/powerpc/mm/book3s64/radix_pgtable.c +0
-5@@ -382,11 +382,6 @@ static void __init radix_init_pgtable(void)*/ register_process_table (__pa (process_tb),0 ,PRTB_SIZE_SHIFT -12 );pr_info ("Process table %p and radix root for kernel: %p\n" , process_tb, init_mm.pgd); - asmvolatile ("ptesync" : : :"memory" ); - asm volatile(PPC_TLBIE_5(%0,%1,2,1,1) : : - "r" (TLBIEL_INVAL_SET_LPID), "r" (0)); - asmvolatile ("eieio; tlbsync; ptesync" : : :"memory" ); -trace_tlbie (0 ,0 ,TLBIEL_INVAL_SET_LPID ,0 ,2 ,1 ,1 );/* * The init_mm context is given the first available (non-zero) PID,
arch/powerpc/mm/book3s64/radix_pgtable.c void __init radix__early_init_mmu(void) · line 632
Quoted 7 lines
arch/powerpc/mm/book3s64/radix_pgtable.c +1
-2@@ -633,8 +628,7 @@ void __init radix__early_init_mmu(void)radix_init_pgtable ();/* Switch to the guard PID before turning on MMU */ radix__switch_mmu_context (NULL , &init_mm); -if (cpu_has_feature (CPU_FTR_HVMODE )) -tlbiel_all (); +tlbiel_all (); }
This is oopsing for me in a guest on Power9:
[ 0.000000] radix-mmu: Page sizes from device-tree: [ 0.000000] radix-mmu: Page size shift = 12 AP=0x0 [ 0.000000] radix-mmu: Page size shift = 16 AP=0x5 [ 0.000000] radix-mmu: Page size shift = 21 AP=0x1 [ 0.000000] radix-mmu: Page size shift = 30 AP=0x2 [ 0.000000] -> fw_vec5_feature_init() [ 0.000000] <- fw_vec5_feature_init() [ 0.000000] -> fw_hypertas_feature_init() [ 0.000000] <- fw_hypertas_feature_init() [ 0.000000] radix-mmu: Activating Kernel Userspace Execution Prevention [ 0.000000] radix-mmu: Activating Kernel Userspace Access Prevention [ 0.000000] lpar: Using radix MMU under hypervisor [ 0.000000] radix-mmu: Mapped 0x0000000000000000-0x0000000040000000 with 1.00 GiB pages (exec) [ 0.000000] radix-mmu: Mapped 0x0000000040000000-0x0000000100000000 with 1.00 GiB pages [ 0.000000] radix-mmu: Process table (____ptrval____) and radix root for kernel: (____ptrval____) [ 0.000000] Oops: Exception in kernel mode, sig: 4 [#1] [ 0.000000] LE PAGE_SIZEdK MMU=Radix MMU=Hash SMP NR_CPUS 48 NUMA [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 5.3.0-rc2-gcc-8.2.0-00063-gef906dcf7b75 #633 [ 0.000000] NIP: c0000000000838f8 LR: c000000001066864 CTR: c0000000000838c0 [ 0.000000] REGS: c000000001647c40 TRAP: 0700 Not tainted (5.3.0-rc2-gcc-8.2.0-00063-gef906dcf7b75) [ 0.000000] MSR: 8000000000043003 <SF,FP,ME,RI,LE> CR: 48000222 XER: 20040000 [ 0.000000] CFAR: c0000000000839b4 IRQMASK: 1 [ 0.000000] GPR00: c000000001066864 c000000001647ed0 c000000001649700 0000000000000000 [ 0.000000] GPR04: c000000001608830 0000000000000000 0000000000000010 2000000000000000 [ 0.000000] GPR08: 0000000000000c00 0000000000000000 0000000000000002 726f6620746f6f72 [ 0.000000] GPR12: c0000000000838c0 c000000001930000 000000000dc5bef0 0000000001309e10 [ 0.000000] GPR16: 0000000001309c90 fffffffffffffffd 000000000dc5bef0 0000000001339800 [ 0.000000] GPR20: 0000000000000014 0000000001ac0000 000000000dc5bf38 000000000daf0000 [ 0.000000] GPR24: 0000000001f4000c c000000000000000 0000000000400000 c000000001802858 [ 0.000000] GPR28: c007ffffffffffff c000000001803954 c000000001681cb0 c000000001608830 [ 0.000000] NIP [c0000000000838f8] radix__tlbiel_all+0x48/0x110 [ 0.000000] LR [c000000001066864] radix__early_init_mmu+0x494/0x4c8 [ 0.000000] Call Trace: [ 0.000000] [c000000001647ed0] [c000000001066820] radix__early_init_mmu+0x450/0x4c8 (unreliable) [ 0.000000] [c000000001647f60] [c00000000105c628] early_setup+0x160/0x198 [ 0.000000] [c000000001647f90] [000000000000b460] 0xb460 [ 0.000000] Instruction dump: [ 0.000000] 2b830001 39000002 409e00e8 3d220003 3929c318 e9290000 e9290010 75290002 [ 0.000000] 41820088 7c4004ac 39200000 79085564 <7d294224> 3940007f 39201000 38e00000 [ 0.000000] random: get_random_bytes called from print_oops_end_marker+0x40/0x80 with crng_init=0 [ 0.000000] ---[ end trace 0000000000000000 ]---
So I think we still need a HV check in there somewhere.
cheers
Nicholas Piggin
reviewerRe: [PATCH 2/3] powerpc/64s/radix: all CPUs should flush local translation structure before turning
Michael Ellerman's on August 19, 2019 12:00 pm:
Quoted 84 lines
Quoted 29 lines · Nicholas Piggin
Rather than sprinkle various translation structure invalidations around different places in early boot, have each CPU flush everything from its local translation structures before enabling its MMU.
Radix guests can execute tlbie(l), so have them
tlbiel_allin the same place as radix host does.Signed-off-by Nicholas Piggin <[email protected]>--- arch/powerpc/mm/book3s64/radix_pgtable.c | 11 ++--------- 1 file changed, 2 insertions(+), 9 deletions(-)arch/powerpc/mm/book3s64/radix_pgtable.c +0
-5@@ -382,11 +382,6 @@ static void __init radix_init_pgtable(void)*/ register_process_table (__pa (process_tb),0 ,PRTB_SIZE_SHIFT -12 );pr_info ("Process table %p and radix root for kernel: %p\n" , process_tb, init_mm.pgd); - asmvolatile ("ptesync" : : :"memory" ); - asm volatile(PPC_TLBIE_5(%0,%1,2,1,1) : : - "r" (TLBIEL_INVAL_SET_LPID), "r" (0)); - asmvolatile ("eieio; tlbsync; ptesync" : : :"memory" ); -trace_tlbie (0 ,0 ,TLBIEL_INVAL_SET_LPID ,0 ,2 ,1 ,1 );/* * The init_mm context is given the first available (non-zero) PID, arch/powerpc/mm/book3s64/radix_pgtable.c void __init radix__early_init_mmu(void) · line 632
Quoted 7 lines
arch/powerpc/mm/book3s64/radix_pgtable.c +1
-2@@ -633,8 +628,7 @@ void __init radix__early_init_mmu(void)radix_init_pgtable ();/* Switch to the guard PID before turning on MMU */ radix__switch_mmu_context (NULL , &init_mm); -if (cpu_has_feature (CPU_FTR_HVMODE )) -tlbiel_all (); +tlbiel_all (); }This is oopsing for me in a guest on Power9:
[ 0.000000] radix-mmu: Page sizes from device-tree: [ 0.000000] radix-mmu: Page size shift = 12 AP=0x0 [ 0.000000] radix-mmu: Page size shift = 16 AP=0x5 [ 0.000000] radix-mmu: Page size shift = 21 AP=0x1 [ 0.000000] radix-mmu: Page size shift = 30 AP=0x2 [ 0.000000] -> fw_vec5_feature_init() [ 0.000000] <- fw_vec5_feature_init() [ 0.000000] -> fw_hypertas_feature_init() [ 0.000000] <- fw_hypertas_feature_init() [ 0.000000] radix-mmu: Activating Kernel Userspace Execution Prevention [ 0.000000] radix-mmu: Activating Kernel Userspace Access Prevention [ 0.000000] lpar: Using radix MMU under hypervisor [ 0.000000] radix-mmu: Mapped 0x0000000000000000-0x0000000040000000 with 1.00 GiB pages (exec) [ 0.000000] radix-mmu: Mapped 0x0000000040000000-0x0000000100000000 with 1.00 GiB pages [ 0.000000] radix-mmu: Process table (____ptrval____) and radix root for kernel: (____ptrval____) [ 0.000000] Oops: Exception in kernel mode, sig: 4 [#1] [ 0.000000] LE PAGE_SIZE=64K MMU=Radix MMU=Hash SMP NR_CPUS=2048 NUMA [ 0.000000] Modules linked in: [ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 5.3.0-rc2-gcc-8.2.0-00063-gef906dcf7b75 #633 [ 0.000000] NIP: c0000000000838f8 LR: c000000001066864 CTR: c0000000000838c0 [ 0.000000] REGS: c000000001647c40 TRAP: 0700 Not tainted (5.3.0-rc2-gcc-8.2.0-00063-gef906dcf7b75) [ 0.000000] MSR: 8000000000043003 <SF,FP,ME,RI,LE> CR: 48000222 XER: 20040000 [ 0.000000] CFAR: c0000000000839b4 IRQMASK: 1 [ 0.000000] GPR00: c000000001066864 c000000001647ed0 c000000001649700 0000000000000000 [ 0.000000] GPR04: c000000001608830 0000000000000000 0000000000000010 2000000000000000 [ 0.000000] GPR08: 0000000000000c00 0000000000000000 0000000000000002 726f6620746f6f72 [ 0.000000] GPR12: c0000000000838c0 c000000001930000 000000000dc5bef0 0000000001309e10 [ 0.000000] GPR16: 0000000001309c90 fffffffffffffffd 000000000dc5bef0 0000000001339800 [ 0.000000] GPR20: 0000000000000014 0000000001ac0000 000000000dc5bf38 000000000daf0000 [ 0.000000] GPR24: 0000000001f4000c c000000000000000 0000000000400000 c000000001802858 [ 0.000000] GPR28: c007ffffffffffff c000000001803954 c000000001681cb0 c000000001608830 [ 0.000000] NIP [c0000000000838f8] radix__tlbiel_all+0x48/0x110 [ 0.000000] LR [c000000001066864] radix__early_init_mmu+0x494/0x4c8 [ 0.000000] Call Trace: [ 0.000000] [c000000001647ed0] [c000000001066820] radix__early_init_mmu+0x450/0x4c8 (unreliable) [ 0.000000] [c000000001647f60] [c00000000105c628] early_setup+0x160/0x198 [ 0.000000] [c000000001647f90] [000000000000b460] 0xb460 [ 0.000000] Instruction dump: [ 0.000000] 2b830001 39000002 409e00e8 3d220003 3929c318 e9290000 e9290010 75290002 [ 0.000000] 41820088 7c4004ac 39200000 79085564 <7d294224> 3940007f 39201000 38e00000 [ 0.000000] random: get_random_bytes called from print_oops_end_marker+0x40/0x80 with crng_init=0 [ 0.000000] ---[ end trace 0000000000000000 ]---
Quoted 1 line
So I think we still need a HV check in there somewhere.
Oh it's the partition scope flush by the looks. Simulator doesn't seem to be trapping that one.
I'll resend, have to fix the kernel invalidation to also do tlbie for nest.
Thanks,
Nick
Nicholas Piggin
reviewerPATCH 3/3 powerpc/64s: introduce options to disable use of the tlbie instruction
In reply to Nicholas PigginIntroduce two options to control the use of the tlbie instruction. A boot time option which completely disables the kernel using the instruction, this is currently incompatible with HASH MMU, KVM, and coherent accelerators.
And a debugfs option can be switched at runtime and avoids using tlbie for invalidating CPU TLBs for normal process and kernel address mappings. Coherent accelerators are still managed with tlbie, as will KVM partition scope translations.
Cross-CPU TLB flushing is implemented with IPIs and tlbiel. This is a basic implementation which does not attempt to make any optimisation beyond the tlbie implementation.
This is useful for performance testing among other things. For example in certain situations on large systems, using IPIs may be faster than tlbie as they can be directed rather than broadcast. Later we may also take advantage of the IPIs to do more interesting things such as trim the mm cpumask more aggressively.
--- .../admin-guide/kernel-parameters.txt | 4 + arch/powerpc/include/asm/book3s/64/tlbflush.h | 9 + arch/powerpc/kvm/book3s_hv.c | 6 + arch/powerpc/mm/book3s64/pgtable.c | 47 +++++ arch/powerpc/mm/book3s64/radix_tlb.c | 182 ++++++++++++++++-- drivers/misc/cxl/main.c | 4 + drivers/misc/ocxl/main.c | 4 + 7 files changed, 238 insertions(+), 18 deletions(-)
Documentation/admin-guide/kernel-parameters.txt +4 -0
@@ -860,6 +860,10 @@ disable_radix [PPC] Disable RADIX MMU mode on POWER9 + disable_tlbie [PPC] + Disable TLBIE instruction. Currently does not work + with KVM, with HASH MMU, or with coherent accelerators. + disable_cpu_apicid= [X86,APIC,SMP] Format: <int> The number of initial APIC ID for the
arch/powerpc/include/asm/book3s/64/tlbflush.h +9 -0
@@ -162,4 +162,13 @@ static inline void flush_tlb_pgtable(struct mmu_gather *tlb, unsigned long addre radix__flush_tlb_pwc(tlb, address); } + +extern bool tlbie_capable; +extern bool tlbie_enabled; + +static inline bool use_tlbie (void ) +{ +return tlbie_enabled; +} +#endif /* _ASM_POWERPC_BOOK3S_64_TLBFLUSH_H */
arch/powerpc/kvm/book3s_hv.c +6 -0
@@ -5462,6 +5462,12 @@ static int kvmppc_radix_possible(void)static int kvmppc_book3s_init_hv (void ) {int r; + +if (!tlbie_capable) { +pr_err ("KVM-HV: Host does not support TLBIE\n" ); +return -ENODEV ; + } +/* * FIXME!! Do we need to check on all cpus ? */
arch/powerpc/mm/book3s64/pgtable.c +47 -0
56 lines not on this page · raw message
arch/powerpc/mm/book3s64/radix_tlb.c +164 -18
273 lines not on this page · raw message
drivers/misc/cxl/main.c +4 -0
@@ -18,6 +18,7 @@#include <linux/sched/task.h> #include <asm/cputable.h> +#include <asm/mmu.h> #include <misc/cxl-base.h> #include "cxl.h"
@@ -315,6 +316,9 @@ static int __init init_cxl(void) {int rc =0 ; +if (!tlbie_capable) +return -EINVAL ; +if ((rc =cxl_file_init ()))return rc;
drivers/misc/ocxl/main.c +4 -0
@@ -2,12 +2,16 @@// Copyright 2017 IBM Corp. #include <linux/module.h> #include <linux/pci.h> +#include <asm/mmu.h> #include "ocxl_internal.h" static int __init init_ocxl (void ) {int rc =0 ; +if (!tlbie_capable) +return -EINVAL ; + rc =ocxl_file_init ();if (rc)return rc;
Signature · 1 line
2.22.0