| Message ID | 20180629213703.13833-1-richard.henderson@linaro.org |
|---|---|
| State | Superseded |
| Headers | show |
| Series | accel/tcg: Avoid caching overwritten tlb entries | expand |
On 29 June 2018 at 22:37, Richard Henderson <richard.henderson@linaro.org> wrote: > When installing a TLB entry, remove any cached version of the > same page in the VTLB. If the existing TLB entry matches, do > not copy into the VTLB, but overwrite it. > > Signed-off-by: Richard Henderson <richard.henderson@linaro.org> > --- > > This may fix some problems with Q800 that Laurent reported. > > On IRC, Peter suggested that regardless of the m68k ptest insn, we > need to be more careful with installed TLB entries. > > I added some temporary logging and concur. This sort of overwrite > happens often when writable pages are marked read-only in order to > track a dirty bit. After the dirty bit is set, we re-install the > TLB entry as read-write. > > I'm mildly surprised we haven't run into problems before... > > > r~ > > --- > accel/tcg/cputlb.c | 60 +++++++++++++++++++++++++--------------------- > 1 file changed, 33 insertions(+), 27 deletions(-) > > diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c > index cc90a5fe92..250b024c5d 100644 > --- a/accel/tcg/cputlb.c > +++ b/accel/tcg/cputlb.c > @@ -235,17 +235,30 @@ void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *src_cpu, > async_safe_run_on_cpu(src_cpu, fn, RUN_ON_CPU_HOST_INT(idxmap)); > } > > - > - > -static inline void tlb_flush_entry(CPUTLBEntry *tlb_entry, target_ulong addr) > +static inline bool tlb_hit_page_anyprot(CPUTLBEntry *tlb_entry, > + target_ulong page) > { > - if (tlb_hit_page(tlb_entry->addr_read, addr) || > - tlb_hit_page(tlb_entry->addr_write, addr) || > - tlb_hit_page(tlb_entry->addr_code, addr)) { > + return (tlb_hit_page(tlb_entry->addr_read, page) || > + tlb_hit_page(tlb_entry->addr_write, page) || > + tlb_hit_page(tlb_entry->addr_code, page)); Checkpatch warns that the outer brackets here are not required. > + /* If the old entry matches the new page, just overwrite TE. */ > + if (!tlb_hit_page_anyprot(te, vaddr_page)) { I found it slightly confusing that the sense of the "if" in the comment is backwards from the "if" in the code. Maybe /* * Only evict the old entry to the victim tlb if it's for a * different page; otherwise just overwrite the stale data. */ ? (I was only slightly confused, though, so I don't feel very strongly here.) > + unsigned vidx = env->vtlb_index++ % CPU_VTLB_SIZE; > + CPUTLBEntry *tv = &env->tlb_v_table[mmu_idx][vidx]; > > - env->iotlb_v[mmu_idx][vidx] = env->iotlb[mmu_idx][index]; > + /* Evict the old entry into the victim tlb. */ > + copy_tlb_helper(tv, te, true); > + env->iotlb_v[mmu_idx][vidx] = env->iotlb[mmu_idx][index]; > + } Otherwise Reviewed-by: Peter Maydell <peter.maydell@linaro.org> thanks -- PMM
On 07/02/2018 03:37 AM, Peter Maydell wrote: > On 29 June 2018 at 22:37, Richard Henderson > <richard.henderson@linaro.org> wrote: >> When installing a TLB entry, remove any cached version of the >> same page in the VTLB. If the existing TLB entry matches, do >> not copy into the VTLB, but overwrite it. >> >> Signed-off-by: Richard Henderson <richard.henderson@linaro.org> >> --- >> >> This may fix some problems with Q800 that Laurent reported. >> >> On IRC, Peter suggested that regardless of the m68k ptest insn, we >> need to be more careful with installed TLB entries. >> >> I added some temporary logging and concur. This sort of overwrite >> happens often when writable pages are marked read-only in order to >> track a dirty bit. After the dirty bit is set, we re-install the >> TLB entry as read-write. >> >> I'm mildly surprised we haven't run into problems before... >> >> >> r~ >> >> --- >> accel/tcg/cputlb.c | 60 +++++++++++++++++++++++++--------------------- >> 1 file changed, 33 insertions(+), 27 deletions(-) >> >> diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c >> index cc90a5fe92..250b024c5d 100644 >> --- a/accel/tcg/cputlb.c >> +++ b/accel/tcg/cputlb.c >> @@ -235,17 +235,30 @@ void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *src_cpu, >> async_safe_run_on_cpu(src_cpu, fn, RUN_ON_CPU_HOST_INT(idxmap)); >> } >> >> - >> - >> -static inline void tlb_flush_entry(CPUTLBEntry *tlb_entry, target_ulong addr) >> +static inline bool tlb_hit_page_anyprot(CPUTLBEntry *tlb_entry, >> + target_ulong page) >> { >> - if (tlb_hit_page(tlb_entry->addr_read, addr) || >> - tlb_hit_page(tlb_entry->addr_write, addr) || >> - tlb_hit_page(tlb_entry->addr_code, addr)) { >> + return (tlb_hit_page(tlb_entry->addr_read, page) || >> + tlb_hit_page(tlb_entry->addr_write, page) || >> + tlb_hit_page(tlb_entry->addr_code, page)); > > Checkpatch warns that the outer brackets here are not required. Yeah, well, emacs requires them for alignment. Checkpatch isn't too smart about multi-line expressions. > > >> + /* If the old entry matches the new page, just overwrite TE. */ >> + if (!tlb_hit_page_anyprot(te, vaddr_page)) { > > I found it slightly confusing that the sense of the "if" in > the comment is backwards from the "if" in the code. > Maybe > /* > * Only evict the old entry to the victim tlb if it's for a > * different page; otherwise just overwrite the stale data. > */ > ? The if got reorganized a few times before settling on the current. I agree that your comment matches the current form much better. r~
diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c index cc90a5fe92..250b024c5d 100644 --- a/accel/tcg/cputlb.c +++ b/accel/tcg/cputlb.c @@ -235,17 +235,30 @@ void tlb_flush_by_mmuidx_all_cpus_synced(CPUState *src_cpu, async_safe_run_on_cpu(src_cpu, fn, RUN_ON_CPU_HOST_INT(idxmap)); } - - -static inline void tlb_flush_entry(CPUTLBEntry *tlb_entry, target_ulong addr) +static inline bool tlb_hit_page_anyprot(CPUTLBEntry *tlb_entry, + target_ulong page) { - if (tlb_hit_page(tlb_entry->addr_read, addr) || - tlb_hit_page(tlb_entry->addr_write, addr) || - tlb_hit_page(tlb_entry->addr_code, addr)) { + return (tlb_hit_page(tlb_entry->addr_read, page) || + tlb_hit_page(tlb_entry->addr_write, page) || + tlb_hit_page(tlb_entry->addr_code, page)); +} + +static inline void tlb_flush_entry(CPUTLBEntry *tlb_entry, target_ulong page) +{ + if (tlb_hit_page_anyprot(tlb_entry, page)) { memset(tlb_entry, -1, sizeof(*tlb_entry)); } } +static inline void tlb_flush_vtlb_page(CPUArchState *env, int mmu_idx, + target_ulong page) +{ + int k; + for (k = 0; k < CPU_VTLB_SIZE; k++) { + tlb_flush_entry(&env->tlb_v_table[mmu_idx][k], page); + } +} + static void tlb_flush_page_async_work(CPUState *cpu, run_on_cpu_data data) { CPUArchState *env = cpu->env_ptr; @@ -271,14 +284,7 @@ static void tlb_flush_page_async_work(CPUState *cpu, run_on_cpu_data data) i = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1); for (mmu_idx = 0; mmu_idx < NB_MMU_MODES; mmu_idx++) { tlb_flush_entry(&env->tlb_table[mmu_idx][i], addr); - } - - /* check whether there are entries that need to be flushed in the vtlb */ - for (mmu_idx = 0; mmu_idx < NB_MMU_MODES; mmu_idx++) { - int k; - for (k = 0; k < CPU_VTLB_SIZE; k++) { - tlb_flush_entry(&env->tlb_v_table[mmu_idx][k], addr); - } + tlb_flush_vtlb_page(env, mmu_idx, addr); } tb_flush_jmp_cache(cpu, addr); @@ -310,7 +316,6 @@ static void tlb_flush_page_by_mmuidx_async_work(CPUState *cpu, unsigned long mmu_idx_bitmap = addr_and_mmuidx & ALL_MMUIDX_BITS; int page = (addr >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1); int mmu_idx; - int i; assert_cpu_is_self(cpu); @@ -320,11 +325,7 @@ static void tlb_flush_page_by_mmuidx_async_work(CPUState *cpu, for (mmu_idx = 0; mmu_idx < NB_MMU_MODES; mmu_idx++) { if (test_bit(mmu_idx, &mmu_idx_bitmap)) { tlb_flush_entry(&env->tlb_table[mmu_idx][page], addr); - - /* check whether there are vltb entries that need to be flushed */ - for (i = 0; i < CPU_VTLB_SIZE; i++) { - tlb_flush_entry(&env->tlb_v_table[mmu_idx][i], addr); - } + tlb_flush_vtlb_page(env, mmu_idx, addr); } } @@ -609,10 +610,9 @@ void tlb_set_page_with_attrs(CPUState *cpu, target_ulong vaddr, target_ulong address; target_ulong code_address; uintptr_t addend; - CPUTLBEntry *te, *tv, tn; + CPUTLBEntry *te, tn; hwaddr iotlb, xlat, sz, paddr_page; target_ulong vaddr_page; - unsigned vidx = env->vtlb_index++ % CPU_VTLB_SIZE; int asidx = cpu_asidx_from_attrs(cpu, attrs); assert_cpu_is_self(cpu); @@ -654,19 +654,25 @@ void tlb_set_page_with_attrs(CPUState *cpu, target_ulong vaddr, addend = (uintptr_t)memory_region_get_ram_ptr(section->mr) + xlat; } + /* Make sure there's no cached translation for the new page. */ + tlb_flush_vtlb_page(env, mmu_idx, vaddr_page); + code_address = address; iotlb = memory_region_section_get_iotlb(cpu, section, vaddr_page, paddr_page, xlat, prot, &address); index = (vaddr_page >> TARGET_PAGE_BITS) & (CPU_TLB_SIZE - 1); te = &env->tlb_table[mmu_idx][index]; - /* do not discard the translation in te, evict it into a victim tlb */ - tv = &env->tlb_v_table[mmu_idx][vidx]; - /* addr_write can race with tlb_reset_dirty_range */ - copy_tlb_helper(tv, te, true); + /* If the old entry matches the new page, just overwrite TE. */ + if (!tlb_hit_page_anyprot(te, vaddr_page)) { + unsigned vidx = env->vtlb_index++ % CPU_VTLB_SIZE; + CPUTLBEntry *tv = &env->tlb_v_table[mmu_idx][vidx]; - env->iotlb_v[mmu_idx][vidx] = env->iotlb[mmu_idx][index]; + /* Evict the old entry into the victim tlb. */ + copy_tlb_helper(tv, te, true); + env->iotlb_v[mmu_idx][vidx] = env->iotlb[mmu_idx][index]; + } /* refill the tlb */ /*
When installing a TLB entry, remove any cached version of the same page in the VTLB. If the existing TLB entry matches, do not copy into the VTLB, but overwrite it. Signed-off-by: Richard Henderson <richard.henderson@linaro.org> --- This may fix some problems with Q800 that Laurent reported. On IRC, Peter suggested that regardless of the m68k ptest insn, we need to be more careful with installed TLB entries. I added some temporary logging and concur. This sort of overwrite happens often when writable pages are marked read-only in order to track a dirty bit. After the dirty bit is set, we re-install the TLB entry as read-write. I'm mildly surprised we haven't run into problems before... r~ --- accel/tcg/cputlb.c | 60 +++++++++++++++++++++++++--------------------- 1 file changed, 33 insertions(+), 27 deletions(-) -- 2.17.1