From 4d2be1267fcfb3a4d2198fd696aec5e3dcbce60e Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Thu, 11 Jun 2009 15:28:09 +0530 Subject: perf_counter, x86: Check old-AMD performance monitoring support AMD supports performance monitoring start from K7 (i.e. family 6), so disable it for earlier AMD CPUs. Signed-off-by: Jaswinder Singh Rajput Cc: Robert Richter Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: <1244714289.6923.0.camel@ht.satnam> Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 275bc142cd5d..3c37c3930ca1 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -1459,6 +1459,10 @@ static int intel_pmu_init(void) static int amd_pmu_init(void) { + /* Performance-monitoring supported from K7 and later: */ + if (boot_cpu_data.x86 < 6) + return -ENODEV; + x86_pmu = amd_pmu; switch (boot_cpu_data.x86) { -- cgit v1.2.3 From f4db43a38f7387c3b19c9565124c06ab0c5d6e9a Mon Sep 17 00:00:00 2001 From: Jaswinder Singh Rajput Date: Sat, 13 Jun 2009 01:06:21 +0530 Subject: perf_counter, x86: Update AMD hw caching related event table All AMD models share the same hw caching related event table. Also complete the table with more events. Signed-off-by: Jaswinder Singh Rajput Cc: Robert Richter Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: <1244835381.2802.2.camel@ht.satnam> Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 36 +++++++++++++++--------------------- 1 file changed, 15 insertions(+), 21 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 3c37c3930ca1..77a59a5566a8 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -389,23 +389,23 @@ static u64 intel_pmu_raw_event(u64 event) return event & CORE_EVNTSEL_MASK; } -static const u64 amd_0f_hw_cache_event_ids +static const u64 amd_hw_cache_event_ids [PERF_COUNT_HW_CACHE_MAX] [PERF_COUNT_HW_CACHE_OP_MAX] [PERF_COUNT_HW_CACHE_RESULT_MAX] = { [ C(L1D) ] = { [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_ACCESS) ] = 0x0040, /* Data Cache Accesses */ + [ C(RESULT_MISS) ] = 0x0041, /* Data Cache Misses */ }, [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_ACCESS) ] = 0x0042, /* Data Cache Refills from L2 */ [ C(RESULT_MISS) ] = 0, }, [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_ACCESS) ] = 0x0267, /* Data Prefetcher :attempts */ + [ C(RESULT_MISS) ] = 0x0167, /* Data Prefetcher :cancelled */ }, }, [ C(L1I ) ] = { @@ -418,17 +418,17 @@ static const u64 amd_0f_hw_cache_event_ids [ C(RESULT_MISS) ] = -1, }, [ C(OP_PREFETCH) ] = { - [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_ACCESS) ] = 0x014B, /* Prefetch Instructions :Load */ [ C(RESULT_MISS) ] = 0, }, }, [ C(LL ) ] = { [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_ACCESS) ] = 0x037D, /* Requests to L2 Cache :IC+DC */ + [ C(RESULT_MISS) ] = 0x037E, /* L2 Cache Misses : IC+DC */ }, [ C(OP_WRITE) ] = { - [ C(RESULT_ACCESS) ] = 0, + [ C(RESULT_ACCESS) ] = 0x017F, /* L2 Fill/Writeback */ [ C(RESULT_MISS) ] = 0, }, [ C(OP_PREFETCH) ] = { @@ -438,8 +438,8 @@ static const u64 amd_0f_hw_cache_event_ids }, [ C(DTLB) ] = { [ C(OP_READ) ] = { - [ C(RESULT_ACCESS) ] = 0, - [ C(RESULT_MISS) ] = 0, + [ C(RESULT_ACCESS) ] = 0x0040, /* Data Cache Accesses */ + [ C(RESULT_MISS) ] = 0x0046, /* L1 DTLB and L2 DLTB Miss */ }, [ C(OP_WRITE) ] = { [ C(RESULT_ACCESS) ] = 0, @@ -1465,16 +1465,10 @@ static int amd_pmu_init(void) x86_pmu = amd_pmu; - switch (boot_cpu_data.x86) { - case 0x0f: - case 0x10: - case 0x11: - memcpy(hw_cache_event_ids, amd_0f_hw_cache_event_ids, - sizeof(hw_cache_event_ids)); + /* Events are common for all AMDs */ + memcpy(hw_cache_event_ids, amd_hw_cache_event_ids, + sizeof(hw_cache_event_ids)); - pr_cont("AMD Family 0f/10/11 events, "); - break; - } return 0; } -- cgit v1.2.3 From 5a6cec3abbdb74244caab68db100825a8c4ac02d Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Fri, 29 May 2009 11:25:09 +0200 Subject: perf_counter, x86: Fix call-chain walking Fix the ptregs variant when we hit user-mode tasks. Cc: Frederic Weisbecker Cc: Pekka Enberg Cc: Arjan van de Ven Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 77a59a5566a8..09d8cb69c3f3 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -1644,7 +1644,9 @@ perf_callchain_user(struct pt_regs *regs, struct perf_callchain_entry *entry) const void __user *fp; int nr = entry->nr; - regs = (struct pt_regs *)current->thread.sp0 - 1; + if (!user_mode(regs)) + regs = task_pt_regs(current); + fp = (void __user *)regs->bp; callchain_store(entry, regs->ip); @@ -1656,7 +1658,7 @@ perf_callchain_user(struct pt_regs *regs, struct perf_callchain_entry *entry) if (!copy_stack_frame(fp, &frame)) break; - if ((unsigned long)fp < user_stack_pointer(regs)) + if ((unsigned long)fp < regs->sp) break; callchain_store(entry, frame.return_address); -- cgit v1.2.3 From 038e836e97e70c4ad2b5058b07fc7207f50b59dd Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 15 Jun 2009 09:57:59 +0200 Subject: perf_counter, x86: Fix kernel-space call-chains Kernel-space call-chains were trimmed at the first entry because we never processed anything beyond the first stack context. Allow the backtrace to jump from NMI to IRQ stack then to task stack and finally user-space stack. Also calculate the stack and bp variables correctly so that the stack walker does not exit early. We can get deep traces as a result, visible in perf report -D output: 0x32af0 [0xe0]: PERF_EVENT (IP, 5): 15134: 0xffffffff815225fd period: 1 ... chain: u:2, k:22, nr:24 ..... 0: 0xffffffff815225fd ..... 1: 0xffffffff810ac51c ..... 2: 0xffffffff81018e29 ..... 3: 0xffffffff81523939 ..... 4: 0xffffffff81524b8f ..... 5: 0xffffffff81524bd9 ..... 6: 0xffffffff8105e498 ..... 7: 0xffffffff8152315a ..... 8: 0xffffffff81522c3a ..... 9: 0xffffffff810d9b74 ..... 10: 0xffffffff810dbeec ..... 11: 0xffffffff810dc3fb This is a 22-entries kernel-space chain. (We still only record reliable stack entries.) Cc: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 22 +++++++++------------- 1 file changed, 9 insertions(+), 13 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 09d8cb69c3f3..6d5e7cfd97e7 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -1575,8 +1575,8 @@ static void backtrace_warning(void *data, char *msg) static int backtrace_stack(void *data, char *name) { - /* Don't bother with IRQ stacks for now */ - return -1; + /* Process all stacks: */ + return 0; } static void backtrace_address(void *data, unsigned long addr, int reliable) @@ -1594,6 +1594,8 @@ static const struct stacktrace_ops backtrace_ops = { .address = backtrace_address, }; +#include "../dumpstack.h" + static void perf_callchain_kernel(struct pt_regs *regs, struct perf_callchain_entry *entry) { @@ -1601,26 +1603,20 @@ perf_callchain_kernel(struct pt_regs *regs, struct perf_callchain_entry *entry) char *stack; int nr = entry->nr; - callchain_store(entry, instruction_pointer(regs)); + callchain_store(entry, regs->ip); stack = ((char *)regs + sizeof(struct pt_regs)); #ifdef CONFIG_FRAME_POINTER - bp = frame_pointer(regs); + get_bp(bp); #else bp = 0; #endif - dump_trace(NULL, regs, (void *)stack, bp, &backtrace_ops, entry); + dump_trace(NULL, regs, (void *)&stack, bp, &backtrace_ops, entry); entry->kernel = entry->nr - nr; } - -struct stack_frame { - const void __user *next_fp; - unsigned long return_address; -}; - static int copy_stack_frame(const void __user *fp, struct stack_frame *frame) { int ret; @@ -1652,7 +1648,7 @@ perf_callchain_user(struct pt_regs *regs, struct perf_callchain_entry *entry) callchain_store(entry, regs->ip); while (entry->nr < MAX_STACK_DEPTH) { - frame.next_fp = NULL; + frame.next_frame = NULL; frame.return_address = 0; if (!copy_stack_frame(fp, &frame)) @@ -1662,7 +1658,7 @@ perf_callchain_user(struct pt_regs *regs, struct perf_callchain_entry *entry) break; callchain_store(entry, frame.return_address); - fp = frame.next_fp; + fp = frame.next_frame; } entry->user = entry->nr - nr; -- cgit v1.2.3 From 465a454f254ee2ff7acc4aececbe31f8af046bc0 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 15 Jun 2009 12:31:37 +0200 Subject: x86, mm: Add __get_user_pages_fast() Introduce a gup_fast() variant which is usable from IRQ/NMI context. Signed-off-by: Peter Zijlstra CC: Nick Piggin Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/mm/gup.c | 56 ++++++++++++++++++++++++++++++++++++++++++++++++++++++ include/linux/mm.h | 6 ++++++ 2 files changed, 62 insertions(+) (limited to 'arch') diff --git a/arch/x86/mm/gup.c b/arch/x86/mm/gup.c index 6340cef6798a..697d5727c119 100644 --- a/arch/x86/mm/gup.c +++ b/arch/x86/mm/gup.c @@ -219,6 +219,62 @@ static int gup_pud_range(pgd_t pgd, unsigned long addr, unsigned long end, return 1; } +/* + * Like get_user_pages_fast() except its IRQ-safe in that it won't fall + * back to the regular GUP. + */ +int __get_user_pages_fast(unsigned long start, int nr_pages, int write, + struct page **pages) +{ + struct mm_struct *mm = current->mm; + unsigned long addr, len, end; + unsigned long next; + unsigned long flags; + pgd_t *pgdp; + int nr = 0; + + start &= PAGE_MASK; + addr = start; + len = (unsigned long) nr_pages << PAGE_SHIFT; + end = start + len; + if (unlikely(!access_ok(write ? VERIFY_WRITE : VERIFY_READ, + (void __user *)start, len))) + return 0; + + /* + * XXX: batch / limit 'nr', to avoid large irq off latency + * needs some instrumenting to determine the common sizes used by + * important workloads (eg. DB2), and whether limiting the batch size + * will decrease performance. + * + * It seems like we're in the clear for the moment. Direct-IO is + * the main guy that batches up lots of get_user_pages, and even + * they are limited to 64-at-a-time which is not so many. + */ + /* + * This doesn't prevent pagetable teardown, but does prevent + * the pagetables and pages from being freed on x86. + * + * So long as we atomically load page table pointers versus teardown + * (which we do on x86, with the above PAE exception), we can follow the + * address down to the the page and take a ref on it. + */ + local_irq_save(flags); + pgdp = pgd_offset(mm, addr); + do { + pgd_t pgd = *pgdp; + + next = pgd_addr_end(addr, end); + if (pgd_none(pgd)) + break; + if (!gup_pud_range(pgd, addr, next, write, pages, &nr)) + break; + } while (pgdp++, addr = next, addr != end); + local_irq_restore(flags); + + return nr; +} + /** * get_user_pages_fast() - pin user pages in memory * @start: starting user address diff --git a/include/linux/mm.h b/include/linux/mm.h index ad613ed66ab0..b457bc047ab1 100644 --- a/include/linux/mm.h +++ b/include/linux/mm.h @@ -862,6 +862,12 @@ extern int mprotect_fixup(struct vm_area_struct *vma, int get_user_pages_fast(unsigned long start, int nr_pages, int write, struct page **pages); +/* + * doesn't attempt to fault and will return short. + */ +int __get_user_pages_fast(unsigned long start, int nr_pages, int write, + struct page **pages); + /* * A callback you can register to apply pressure to ageable caches. * -- cgit v1.2.3 From 3ff0141aa3a03ca3388b40b36167d0a37919f3fd Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 15 Jun 2009 12:40:41 +0200 Subject: x86: Add NMI types for kmap_atomic Two new kmap_atomic slots for NMI context. And teach pte_offset_map() about NMI context. Signed-off-by: Peter Zijlstra CC: Nick Piggin Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Signed-off-by: Ingo Molnar --- arch/x86/include/asm/kmap_types.h | 4 +++- arch/x86/include/asm/pgtable_32.h | 5 +++-- 2 files changed, 6 insertions(+), 3 deletions(-) (limited to 'arch') diff --git a/arch/x86/include/asm/kmap_types.h b/arch/x86/include/asm/kmap_types.h index 5759c165a5cf..ff00a44b7d0d 100644 --- a/arch/x86/include/asm/kmap_types.h +++ b/arch/x86/include/asm/kmap_types.h @@ -21,7 +21,9 @@ D(9) KM_IRQ0, D(10) KM_IRQ1, D(11) KM_SOFTIRQ0, D(12) KM_SOFTIRQ1, -D(13) KM_TYPE_NR +D(13) KM_NMI, +D(14) KM_NMI_PTE, +D(15) KM_TYPE_NR }; #undef D diff --git a/arch/x86/include/asm/pgtable_32.h b/arch/x86/include/asm/pgtable_32.h index 31bd120cf2a2..85464971bca0 100644 --- a/arch/x86/include/asm/pgtable_32.h +++ b/arch/x86/include/asm/pgtable_32.h @@ -49,13 +49,14 @@ extern void set_pmd_pfn(unsigned long, unsigned long, pgprot_t); #endif #if defined(CONFIG_HIGHPTE) +#define __KM_PTE (in_nmi() ? KM_NMI_PTE : KM_PTE0) #define pte_offset_map(dir, address) \ - ((pte_t *)kmap_atomic_pte(pmd_page(*(dir)), KM_PTE0) + \ + ((pte_t *)kmap_atomic_pte(pmd_page(*(dir)), __KM_PTE) + \ pte_index((address))) #define pte_offset_map_nested(dir, address) \ ((pte_t *)kmap_atomic_pte(pmd_page(*(dir)), KM_PTE1) + \ pte_index((address))) -#define pte_unmap(pte) kunmap_atomic((pte), KM_PTE0) +#define pte_unmap(pte) kunmap_atomic((pte), __KM_PTE) #define pte_unmap_nested(pte) kunmap_atomic((pte), KM_PTE1) #else #define pte_offset_map(dir, address) \ -- cgit v1.2.3 From 74193ef0ecab92535c8517f082f1f50504526c9b Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 15 Jun 2009 13:07:24 +0200 Subject: perf_counter: x86: Fix call-chain support to use NMI-safe methods __copy_from_user_inatomic() isn't NMI safe in that it can trigger the page fault handler which is another trap and its return path invokes IRET which will also close the NMI context. Therefore use a GUP based approach to copy the stack frames over. We tried an alternative solution as well: we used a forward ported version of Mathieu Desnoyers's "NMI safe INT3 and Page Fault" patch that modifies the exception return path to use an open-coded IRET with explicit stack unrolling and TF checking. This didnt work as it interacted with faulting user-space instructions, causing them not to restart properly, which corrupts user-space registers. Solving that would probably involve disassembling those instructions and backtracing the RIP. But even without that, the code was deemed rather complex to the already non-trivial x86 entry assembly code, so instead we went for this GUP based method that does a software-walk of the pagetables. Signed-off-by: Peter Zijlstra Cc: Nick Piggin Cc: Pekka Enberg Cc: Vegard Nossum Cc: Jeremy Fitzhardinge Cc: Mathieu Desnoyers Cc: Linus Torvalds Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 49 ++++++++++++++++++++++++++++++-------- 1 file changed, 39 insertions(+), 10 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index 6d5e7cfd97e7..e8c68a5091df 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -19,6 +19,7 @@ #include #include #include +#include #include #include @@ -1617,20 +1618,48 @@ perf_callchain_kernel(struct pt_regs *regs, struct perf_callchain_entry *entry) entry->kernel = entry->nr - nr; } -static int copy_stack_frame(const void __user *fp, struct stack_frame *frame) +/* + * best effort, GUP based copy_from_user() that assumes IRQ or NMI context + */ +static unsigned long +copy_from_user_nmi(void *to, const void __user *from, unsigned long n) { + unsigned long offset, addr = (unsigned long)from; + int type = in_nmi() ? KM_NMI : KM_IRQ0; + unsigned long size, len = 0; + struct page *page; + void *map; int ret; - if (!access_ok(VERIFY_READ, fp, sizeof(*frame))) - return 0; + do { + ret = __get_user_pages_fast(addr, 1, 0, &page); + if (!ret) + break; - ret = 1; - pagefault_disable(); - if (__copy_from_user_inatomic(frame, fp, sizeof(*frame))) - ret = 0; - pagefault_enable(); + offset = addr & (PAGE_SIZE - 1); + size = min(PAGE_SIZE - offset, n - len); - return ret; + map = kmap_atomic(page, type); + memcpy(to, map+offset, size); + kunmap_atomic(map, type); + put_page(page); + + len += size; + to += size; + addr += size; + + } while (len < n); + + return len; +} + +static int copy_stack_frame(const void __user *fp, struct stack_frame *frame) +{ + unsigned long bytes; + + bytes = copy_from_user_nmi(frame, fp, sizeof(*frame)); + + return bytes == sizeof(*frame); } static void @@ -1643,7 +1672,7 @@ perf_callchain_user(struct pt_regs *regs, struct perf_callchain_entry *entry) if (!user_mode(regs)) regs = task_pt_regs(current); - fp = (void __user *)regs->bp; + fp = (void __user *)regs->bp; callchain_store(entry, regs->ip); -- cgit v1.2.3 From 90c8f954534ba15e4542ab00dd9f0e58b071518c Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Mon, 15 Jun 2009 21:36:52 +1000 Subject: perf_counter: powerpc: Fix two compile warnings This fixes a couple of compile warnings that crept into the powerpc perf_counter code recently: CC arch/powerpc/kernel/perf_counter.o arch/powerpc/kernel/perf_counter.c: In function 'record_and_restart': arch/powerpc/kernel/perf_counter.c:1016: warning: unused variable 'addr' arch/powerpc/kernel/perf_counter.c: In function 'hw_perf_counter_init': arch/powerpc/kernel/perf_counter.c:891: warning: 'ev' may be used uninitialized in this function Stephen Rothwell reported this against linux-next as well. Reported-by: Stephen Rothwell Signed-off-by: Paul Mackerras Cc: Peter Zijlstra LKML-Reference: <18998.12884.787039.22202@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- arch/powerpc/kernel/perf_counter.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/powerpc/kernel/perf_counter.c b/arch/powerpc/kernel/perf_counter.c index bb202388170e..e6dc1850191c 100644 --- a/arch/powerpc/kernel/perf_counter.c +++ b/arch/powerpc/kernel/perf_counter.c @@ -913,6 +913,8 @@ const struct pmu *hw_perf_counter_init(struct perf_counter *counter) case PERF_TYPE_RAW: ev = counter->attr.config; break; + default: + return ERR_PTR(-EINVAL); } counter->hw.config_base = ev; counter->hw.idx = 0; @@ -1013,7 +1015,7 @@ static void record_and_restart(struct perf_counter *counter, long val, u64 period = counter->hw.sample_period; s64 prev, delta, left; int record = 0; - u64 addr, mmcra, sdsync; + u64 mmcra, sdsync; /* we don't have to worry about interrupts here */ prev = atomic64_read(&counter->hw.prev_count); -- cgit v1.2.3 From 9974458e2f9a11dbd2f4bd14fab5a79af4907b41 Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Mon, 15 Jun 2009 21:45:16 +1000 Subject: perf_counter: Make set_perf_counter_pending() declaration common At present, every architecture that supports perf_counters has to declare set_perf_counter_pending() in its arch-specific headers. This consolidates the declarations into a single declaration in one common place, include/linux/perf_counter.h. On powerpc, we continue to provide a static inline definition of set_perf_counter_pending() in the powerpc hw_irq.h. Also, this removes from the x86 perf_counter.h the unused null definitions of {test,clear}_perf_counter_pending. Reported-by: Mike Frysinger Signed-off-by: Paul Mackerras Cc: Peter Zijlstra Cc: benh@kernel.crashing.org LKML-Reference: <18998.13388.920691.523227@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- arch/powerpc/include/asm/hw_irq.h | 1 - arch/powerpc/include/asm/perf_counter.h | 2 ++ arch/x86/include/asm/perf_counter.h | 5 ----- include/linux/perf_counter.h | 1 + 4 files changed, 3 insertions(+), 6 deletions(-) (limited to 'arch') diff --git a/arch/powerpc/include/asm/hw_irq.h b/arch/powerpc/include/asm/hw_irq.h index 53512374e1c9..1974cf191b03 100644 --- a/arch/powerpc/include/asm/hw_irq.h +++ b/arch/powerpc/include/asm/hw_irq.h @@ -163,7 +163,6 @@ static inline unsigned long test_perf_counter_pending(void) return 0; } -static inline void set_perf_counter_pending(void) {} static inline void clear_perf_counter_pending(void) {} #endif /* CONFIG_PERF_COUNTERS */ diff --git a/arch/powerpc/include/asm/perf_counter.h b/arch/powerpc/include/asm/perf_counter.h index cc7c887705b8..b398a84edced 100644 --- a/arch/powerpc/include/asm/perf_counter.h +++ b/arch/powerpc/include/asm/perf_counter.h @@ -10,6 +10,8 @@ */ #include +#include + #define MAX_HWCOUNTERS 8 #define MAX_EVENT_ALTERNATIVES 8 #define MAX_LIMITED_HWCOUNTERS 2 diff --git a/arch/x86/include/asm/perf_counter.h b/arch/x86/include/asm/perf_counter.h index 876ed97147b3..5fb33e160ea0 100644 --- a/arch/x86/include/asm/perf_counter.h +++ b/arch/x86/include/asm/perf_counter.h @@ -84,11 +84,6 @@ union cpuid10_edx { #define MSR_ARCH_PERFMON_FIXED_CTR2 0x30b #define X86_PMC_IDX_FIXED_BUS_CYCLES (X86_PMC_IDX_FIXED + 2) -extern void set_perf_counter_pending(void); - -#define clear_perf_counter_pending() do { } while (0) -#define test_perf_counter_pending() (0) - #ifdef CONFIG_PERF_COUNTERS extern void init_hw_perf_counters(void); extern void perf_counters_lapic_init(void); diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h index 1b3118a1023a..eccae437fe37 100644 --- a/include/linux/perf_counter.h +++ b/include/linux/perf_counter.h @@ -604,6 +604,7 @@ extern void perf_counter_task_tick(struct task_struct *task, int cpu); extern int perf_counter_init_task(struct task_struct *child); extern void perf_counter_exit_task(struct task_struct *child); extern void perf_counter_free_task(struct task_struct *task); +extern void set_perf_counter_pending(void); extern void perf_counter_do_pending(void); extern void perf_counter_print_debug(void); extern void __perf_disable(void); -- cgit v1.2.3 From 0990b1c65729012a63e0eeca93aaaafea4e9a064 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 15 Jun 2009 16:46:05 +0200 Subject: x86: Add NMI types for kmap_atomic, fix I just realized this has a kmap_atomic bug in... The below would fix it - but it's complicating this code some more. Alternatively I would have to introduce something like pte_offset_map_irq() which would make the irq/nmi detection and leave the regular code paths alone, however that would mean either duplicating the gup_fast() pagewalk or passing down a pte function pointer, which would only duplicate the gup_pte_range() bit, neither is really attractive ... Signed-off-by: Peter Zijlstra CC: Nick Piggin Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo Signed-off-by: Ingo Molnar --- arch/x86/include/asm/kmap_types.h | 11 ++++++----- arch/x86/include/asm/pgtable_32.h | 5 ++++- 2 files changed, 10 insertions(+), 6 deletions(-) (limited to 'arch') diff --git a/arch/x86/include/asm/kmap_types.h b/arch/x86/include/asm/kmap_types.h index ff00a44b7d0d..f86613846198 100644 --- a/arch/x86/include/asm/kmap_types.h +++ b/arch/x86/include/asm/kmap_types.h @@ -19,11 +19,12 @@ D(7) KM_PTE0, D(8) KM_PTE1, D(9) KM_IRQ0, D(10) KM_IRQ1, -D(11) KM_SOFTIRQ0, -D(12) KM_SOFTIRQ1, -D(13) KM_NMI, -D(14) KM_NMI_PTE, -D(15) KM_TYPE_NR +D(11) KM_IRQ_PTE, +D(12) KM_SOFTIRQ0, +D(13) KM_SOFTIRQ1, +D(14) KM_NMI, +D(15) KM_NMI_PTE, +D(16) KM_TYPE_NR }; #undef D diff --git a/arch/x86/include/asm/pgtable_32.h b/arch/x86/include/asm/pgtable_32.h index 85464971bca0..01fd9461d323 100644 --- a/arch/x86/include/asm/pgtable_32.h +++ b/arch/x86/include/asm/pgtable_32.h @@ -49,7 +49,10 @@ extern void set_pmd_pfn(unsigned long, unsigned long, pgprot_t); #endif #if defined(CONFIG_HIGHPTE) -#define __KM_PTE (in_nmi() ? KM_NMI_PTE : KM_PTE0) +#define __KM_PTE \ + (in_nmi() ? KM_NMI_PTE : \ + in_irq() ? KM_IRQ_PTE : \ + KM_PTE0) #define pte_offset_map(dir, address) \ ((pte_t *)kmap_atomic_pte(pmd_page(*(dir)), __KM_PTE) + \ pte_index((address))) -- cgit v1.2.3 From 60f916dee612130c9977a8edd4abee98334202ba Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Mon, 15 Jun 2009 19:00:20 +0200 Subject: perf_counter: x86: Set the period in the intel overflow handler Commit 9e350de37ac960 ("perf_counter: Accurate period data") missed a spot, which caused all Intel-PMU samples to have a period of 0. This broke auto-freq sampling. Signed-off-by: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 2 ++ 1 file changed, 2 insertions(+) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index e8c68a5091df..ce1ae3f1f86c 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -1224,6 +1224,8 @@ again: if (!intel_pmu_save_and_restart(counter)) continue; + data.period = counter->hw.last_period; + if (perf_counter_overflow(counter, 1, &data)) intel_pmu_disable_counter(&counter->hw, bit); } -- cgit v1.2.3 From 105988c015943e77092a6568bc5fb7e386df6ccd Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Wed, 17 Jun 2009 21:50:04 +1000 Subject: perf_counter: powerpc: Enable use of software counters on 32-bit powerpc This enables the perf_counter subsystem on 32-bit powerpc. Since we don't have any support for hardware counters on 32-bit powerpc yet, only software counters can be used. Besides selecting HAVE_PERF_COUNTERS for 32-bit powerpc as well as 64-bit, the main thing this does is add an implementation of set_perf_counter_pending(). This needs to arrange for perf_counter_do_pending() to be called when interrupts are enabled. Rather than add code to local_irq_restore as 64-bit does, the 32-bit set_perf_counter_pending() generates an interrupt by setting the decrementer to 1 so that a decrementer interrupt will become pending in 1 or 2 timebase ticks (if a decrementer interrupt isn't already pending). When interrupts are enabled, timer_interrupt() will be called, and some new code in there calls perf_counter_do_pending(). We use a per-cpu array of flags to indicate whether we need to call perf_counter_do_pending() or not. This introduces a couple of new Kconfig symbols: PPC_HAVE_PMU_SUPPORT, which is selected by processor families for which we have hardware PMU support (currently only PPC64), and PPC_PERF_CTRS, which enables the powerpc-specific perf_counter back-end. Signed-off-by: Paul Mackerras Cc: Peter Zijlstra Cc: linuxppc-dev@ozlabs.org Cc: benh@kernel.crashing.org LKML-Reference: <19000.55404.103840.393470@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- arch/powerpc/Kconfig | 1 + arch/powerpc/include/asm/hw_irq.h | 5 ++++- arch/powerpc/include/asm/perf_counter.h | 10 ++++++++-- arch/powerpc/kernel/Makefile | 6 +++--- arch/powerpc/kernel/time.c | 25 +++++++++++++++++++++++++ arch/powerpc/platforms/Kconfig.cputype | 11 ++++++++++- 6 files changed, 51 insertions(+), 7 deletions(-) (limited to 'arch') diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig index 9fb344d5a86a..bf6cedfa05db 100644 --- a/arch/powerpc/Kconfig +++ b/arch/powerpc/Kconfig @@ -126,6 +126,7 @@ config PPC select HAVE_OPROFILE select HAVE_SYSCALL_WRAPPERS if PPC64 select GENERIC_ATOMIC64 if PPC32 + select HAVE_PERF_COUNTERS config EARLY_PRINTK bool diff --git a/arch/powerpc/include/asm/hw_irq.h b/arch/powerpc/include/asm/hw_irq.h index 10a642df014e..867ab8ed69b3 100644 --- a/arch/powerpc/include/asm/hw_irq.h +++ b/arch/powerpc/include/asm/hw_irq.h @@ -131,6 +131,8 @@ static inline int irqs_disabled_flags(unsigned long flags) struct irq_chip; #ifdef CONFIG_PERF_COUNTERS + +#ifdef CONFIG_PPC64 static inline unsigned long test_perf_counter_pending(void) { unsigned long x; @@ -154,8 +156,9 @@ static inline void clear_perf_counter_pending(void) "r" (0), "i" (offsetof(struct paca_struct, perf_counter_pending))); } +#endif /* CONFIG_PPC64 */ -#else +#else /* CONFIG_PERF_COUNTERS */ static inline unsigned long test_perf_counter_pending(void) { diff --git a/arch/powerpc/include/asm/perf_counter.h b/arch/powerpc/include/asm/perf_counter.h index b398a84edced..2c2d9f643df0 100644 --- a/arch/powerpc/include/asm/perf_counter.h +++ b/arch/powerpc/include/asm/perf_counter.h @@ -57,10 +57,16 @@ extern struct power_pmu *ppmu; struct pt_regs; extern unsigned long perf_misc_flags(struct pt_regs *regs); -#define perf_misc_flags(regs) perf_misc_flags(regs) - extern unsigned long perf_instruction_pointer(struct pt_regs *regs); +/* + * Only override the default definitions in include/linux/perf_counter.h + * if we have hardware PMU support. + */ +#ifdef CONFIG_PPC_PERF_CTRS +#define perf_misc_flags(regs) perf_misc_flags(regs) +#endif + /* * The power_pmu.get_constraint function returns a 64-bit value and * a 64-bit mask that express the constraints between this event and diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile index 612b0c4dc26d..c5f93f061927 100644 --- a/arch/powerpc/kernel/Makefile +++ b/arch/powerpc/kernel/Makefile @@ -95,9 +95,9 @@ obj64-$(CONFIG_AUDIT) += compat_audit.o obj-$(CONFIG_DYNAMIC_FTRACE) += ftrace.o obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o -obj-$(CONFIG_PERF_COUNTERS) += perf_counter.o power4-pmu.o ppc970-pmu.o \ - power5-pmu.o power5+-pmu.o power6-pmu.o \ - power7-pmu.o +obj-$(CONFIG_PPC_PERF_CTRS) += perf_counter.o +obj64-$(CONFIG_PPC_PERF_CTRS) += power4-pmu.o ppc970-pmu.o power5-pmu.o \ + power5+-pmu.o power6-pmu.o power7-pmu.o obj-$(CONFIG_8XX_MINIMAL_FPEMU) += softemu8xx.o diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c index 15391c2ab013..eae4511ceeac 100644 --- a/arch/powerpc/kernel/time.c +++ b/arch/powerpc/kernel/time.c @@ -53,6 +53,7 @@ #include #include #include +#include #include #include @@ -525,6 +526,26 @@ void __init iSeries_time_init_early(void) } #endif /* CONFIG_PPC_ISERIES */ +#if defined(CONFIG_PERF_COUNTERS) && defined(CONFIG_PPC32) +DEFINE_PER_CPU(u8, perf_counter_pending); + +void set_perf_counter_pending(void) +{ + get_cpu_var(perf_counter_pending) = 1; + set_dec(1); + put_cpu_var(perf_counter_pending); +} + +#define test_perf_counter_pending() __get_cpu_var(perf_counter_pending) +#define clear_perf_counter_pending() __get_cpu_var(perf_counter_pending) = 0 + +#else /* CONFIG_PERF_COUNTERS && CONFIG_PPC32 */ + +#define test_perf_counter_pending() 0 +#define clear_perf_counter_pending() + +#endif /* CONFIG_PERF_COUNTERS && CONFIG_PPC32 */ + /* * For iSeries shared processors, we have to let the hypervisor * set the hardware decrementer. We set a virtual decrementer @@ -551,6 +572,10 @@ void timer_interrupt(struct pt_regs * regs) set_dec(DECREMENTER_MAX); #ifdef CONFIG_PPC32 + if (test_perf_counter_pending()) { + clear_perf_counter_pending(); + perf_counter_do_pending(); + } if (atomic_read(&ppc_n_lost_interrupts) != 0) do_IRQ(regs); #endif diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype index cca6b4fc719a..dd9f3ec5ee30 100644 --- a/arch/powerpc/platforms/Kconfig.cputype +++ b/arch/powerpc/platforms/Kconfig.cputype @@ -1,7 +1,7 @@ config PPC64 bool "64-bit kernel" default n - select HAVE_PERF_COUNTERS + select PPC_HAVE_PMU_SUPPORT help This option selects whether a 32-bit or a 64-bit kernel will be built. @@ -243,6 +243,15 @@ config VIRT_CPU_ACCOUNTING If in doubt, say Y here. +config PPC_HAVE_PMU_SUPPORT + bool + +config PPC_PERF_CTRS + def_bool y + depends on PERF_COUNTERS && PPC_HAVE_PMU_SUPPORT + help + This enables the powerpc-specific perf_counter back-end. + config SMP depends on PPC_STD_MMU || FSL_BOOKE bool "Symmetric multi-processing support" -- cgit v1.2.3 From 448d64f8f4c147db466c549550767cc515a4d34c Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Wed, 17 Jun 2009 21:51:13 +1000 Subject: perf_counter: powerpc: Use unsigned long for register and constraint values This changes the powerpc perf_counter back-end to use unsigned long types for hardware register values and for the value/mask pairs used in checking whether a given set of events fit within the hardware constraints. This is in preparation for adding support for the PMU on some 32-bit powerpc processors. On 32-bit processors the hardware registers are only 32 bits wide, and the PMU structure is generally simpler, so 32 bits should be ample for expressing the hardware constraints. On 64-bit processors, unsigned long is 64 bits wide, so using unsigned long vs. u64 (unsigned long long) makes no actual difference. This makes some other very minor changes: adjusting whitespace to line things up in initialized structures, and simplifying some code in hw_perf_disable(). Signed-off-by: Paul Mackerras Cc: Peter Zijlstra Cc: linuxppc-dev@ozlabs.org Cc: benh@kernel.crashing.org LKML-Reference: <19000.55473.26174.331511@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- arch/powerpc/include/asm/perf_counter.h | 35 +++++++------- arch/powerpc/kernel/perf_counter.c | 20 ++++---- arch/powerpc/kernel/power4-pmu.c | 74 +++++++++++++++-------------- arch/powerpc/kernel/power5+-pmu.c | 79 ++++++++++++++++--------------- arch/powerpc/kernel/power5-pmu.c | 83 +++++++++++++++++---------------- arch/powerpc/kernel/power6-pmu.c | 57 +++++++++++----------- arch/powerpc/kernel/power7-pmu.c | 46 +++++++++--------- arch/powerpc/kernel/ppc970-pmu.c | 47 ++++++++++--------- 8 files changed, 229 insertions(+), 212 deletions(-) (limited to 'arch') diff --git a/arch/powerpc/include/asm/perf_counter.h b/arch/powerpc/include/asm/perf_counter.h index 2c2d9f643df0..2ceb0fefa93a 100644 --- a/arch/powerpc/include/asm/perf_counter.h +++ b/arch/powerpc/include/asm/perf_counter.h @@ -21,21 +21,22 @@ * describe the PMU on a particular POWER-family CPU. */ struct power_pmu { - int n_counter; - int max_alternatives; - u64 add_fields; - u64 test_adder; - int (*compute_mmcr)(u64 events[], int n_ev, - unsigned int hwc[], u64 mmcr[]); - int (*get_constraint)(u64 event, u64 *mskp, u64 *valp); - int (*get_alternatives)(u64 event, unsigned int flags, - u64 alt[]); - void (*disable_pmc)(unsigned int pmc, u64 mmcr[]); - int (*limited_pmc_event)(u64 event); - u32 flags; - int n_generic; - int *generic_events; - int (*cache_events)[PERF_COUNT_HW_CACHE_MAX] + int n_counter; + int max_alternatives; + unsigned long add_fields; + unsigned long test_adder; + int (*compute_mmcr)(u64 events[], int n_ev, + unsigned int hwc[], unsigned long mmcr[]); + int (*get_constraint)(u64 event, unsigned long *mskp, + unsigned long *valp); + int (*get_alternatives)(u64 event, unsigned int flags, + u64 alt[]); + void (*disable_pmc)(unsigned int pmc, unsigned long mmcr[]); + int (*limited_pmc_event)(u64 event); + u32 flags; + int n_generic; + int *generic_events; + int (*cache_events)[PERF_COUNT_HW_CACHE_MAX] [PERF_COUNT_HW_CACHE_OP_MAX] [PERF_COUNT_HW_CACHE_RESULT_MAX]; }; @@ -68,8 +69,8 @@ extern unsigned long perf_instruction_pointer(struct pt_regs *regs); #endif /* - * The power_pmu.get_constraint function returns a 64-bit value and - * a 64-bit mask that express the constraints between this event and + * The power_pmu.get_constraint function returns a 32/64-bit value and + * a 32/64-bit mask that express the constraints between this event and * other events. * * The value and mask are divided up into (non-overlapping) bitfields diff --git a/arch/powerpc/kernel/perf_counter.c b/arch/powerpc/kernel/perf_counter.c index e6dc1850191c..9300638b8c26 100644 --- a/arch/powerpc/kernel/perf_counter.c +++ b/arch/powerpc/kernel/perf_counter.c @@ -29,7 +29,7 @@ struct cpu_hw_counters { struct perf_counter *counter[MAX_HWCOUNTERS]; u64 events[MAX_HWCOUNTERS]; unsigned int flags[MAX_HWCOUNTERS]; - u64 mmcr[3]; + unsigned long mmcr[3]; struct perf_counter *limited_counter[MAX_LIMITED_HWCOUNTERS]; u8 limited_hwidx[MAX_LIMITED_HWCOUNTERS]; }; @@ -135,15 +135,15 @@ static void write_pmc(int idx, unsigned long val) static int power_check_constraints(u64 event[], unsigned int cflags[], int n_ev) { - u64 mask, value, nv; + unsigned long mask, value, nv; u64 alternatives[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES]; - u64 amasks[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES]; - u64 avalues[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES]; - u64 smasks[MAX_HWCOUNTERS], svalues[MAX_HWCOUNTERS]; + unsigned long amasks[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES]; + unsigned long avalues[MAX_HWCOUNTERS][MAX_EVENT_ALTERNATIVES]; + unsigned long smasks[MAX_HWCOUNTERS], svalues[MAX_HWCOUNTERS]; int n_alt[MAX_HWCOUNTERS], choice[MAX_HWCOUNTERS]; int i, j; - u64 addf = ppmu->add_fields; - u64 tadd = ppmu->test_adder; + unsigned long addf = ppmu->add_fields; + unsigned long tadd = ppmu->test_adder; if (n_ev > ppmu->n_counter) return -1; @@ -403,14 +403,12 @@ static void write_mmcr0(struct cpu_hw_counters *cpuhw, unsigned long mmcr0) void hw_perf_disable(void) { struct cpu_hw_counters *cpuhw; - unsigned long ret; unsigned long flags; local_irq_save(flags); cpuhw = &__get_cpu_var(cpu_hw_counters); - ret = cpuhw->disabled; - if (!ret) { + if (!cpuhw->disabled) { cpuhw->disabled = 1; cpuhw->n_added = 0; @@ -1013,9 +1011,9 @@ static void record_and_restart(struct perf_counter *counter, long val, struct pt_regs *regs, int nmi) { u64 period = counter->hw.sample_period; + unsigned long mmcra, sdsync; s64 prev, delta, left; int record = 0; - u64 mmcra, sdsync; /* we don't have to worry about interrupts here */ prev = atomic64_read(&counter->hw.prev_count); diff --git a/arch/powerpc/kernel/power4-pmu.c b/arch/powerpc/kernel/power4-pmu.c index 07bd308a5fa7..81a1708f83b2 100644 --- a/arch/powerpc/kernel/power4-pmu.c +++ b/arch/powerpc/kernel/power4-pmu.c @@ -179,22 +179,22 @@ static short mmcr1_adder_bits[8] = { */ static struct unitinfo { - u64 value, mask; - int unit; - int lowerbit; + unsigned long value, mask; + int unit; + int lowerbit; } p4_unitinfo[16] = { - [PM_FPU] = { 0x44000000000000ull, 0x88000000000000ull, PM_FPU, 0 }, - [PM_ISU1] = { 0x20080000000000ull, 0x88000000000000ull, PM_ISU1, 0 }, + [PM_FPU] = { 0x44000000000000ul, 0x88000000000000ul, PM_FPU, 0 }, + [PM_ISU1] = { 0x20080000000000ul, 0x88000000000000ul, PM_ISU1, 0 }, [PM_ISU1_ALT] = - { 0x20080000000000ull, 0x88000000000000ull, PM_ISU1, 0 }, - [PM_IFU] = { 0x02200000000000ull, 0x08820000000000ull, PM_IFU, 41 }, + { 0x20080000000000ul, 0x88000000000000ul, PM_ISU1, 0 }, + [PM_IFU] = { 0x02200000000000ul, 0x08820000000000ul, PM_IFU, 41 }, [PM_IFU_ALT] = - { 0x02200000000000ull, 0x08820000000000ull, PM_IFU, 41 }, - [PM_IDU0] = { 0x10100000000000ull, 0x80840000000000ull, PM_IDU0, 1 }, - [PM_ISU2] = { 0x10140000000000ull, 0x80840000000000ull, PM_ISU2, 0 }, - [PM_LSU0] = { 0x01400000000000ull, 0x08800000000000ull, PM_LSU0, 0 }, - [PM_LSU1] = { 0x00000000000000ull, 0x00010000000000ull, PM_LSU1, 40 }, - [PM_GPS] = { 0x00000000000000ull, 0x00000000000000ull, PM_GPS, 0 } + { 0x02200000000000ul, 0x08820000000000ul, PM_IFU, 41 }, + [PM_IDU0] = { 0x10100000000000ul, 0x80840000000000ul, PM_IDU0, 1 }, + [PM_ISU2] = { 0x10140000000000ul, 0x80840000000000ul, PM_ISU2, 0 }, + [PM_LSU0] = { 0x01400000000000ul, 0x08800000000000ul, PM_LSU0, 0 }, + [PM_LSU1] = { 0x00000000000000ul, 0x00010000000000ul, PM_LSU1, 40 }, + [PM_GPS] = { 0x00000000000000ul, 0x00000000000000ul, PM_GPS, 0 } }; static unsigned char direct_marked_event[8] = { @@ -249,10 +249,11 @@ static int p4_marked_instr_event(u64 event) return (mask >> (byte * 8 + bit)) & 1; } -static int p4_get_constraint(u64 event, u64 *maskp, u64 *valp) +static int p4_get_constraint(u64 event, unsigned long *maskp, + unsigned long *valp) { int pmc, byte, unit, lower, sh; - u64 mask = 0, value = 0; + unsigned long mask = 0, value = 0; int grp = -1; pmc = (event >> PM_PMC_SH) & PM_PMC_MSK; @@ -282,14 +283,14 @@ static int p4_get_constraint(u64 event, u64 *maskp, u64 *valp) value |= p4_unitinfo[unit].value; sh = p4_unitinfo[unit].lowerbit; if (sh > 1) - value |= (u64)lower << sh; + value |= (unsigned long)lower << sh; else if (lower != sh) return -1; unit = p4_unitinfo[unit].unit; /* Set byte lane select field */ mask |= 0xfULL << (28 - 4 * byte); - value |= (u64)unit << (28 - 4 * byte); + value |= (unsigned long)unit << (28 - 4 * byte); } if (grp == 0) { /* increment PMC1/2/5/6 field */ @@ -353,9 +354,9 @@ static int p4_get_alternatives(u64 event, unsigned int flags, u64 alt[]) } static int p4_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], u64 mmcr[]) + unsigned int hwc[], unsigned long mmcr[]) { - u64 mmcr0 = 0, mmcr1 = 0, mmcra = 0; + unsigned long mmcr0 = 0, mmcr1 = 0, mmcra = 0; unsigned int pmc, unit, byte, psel, lower; unsigned int ttm, grp; unsigned int pmc_inuse = 0; @@ -429,9 +430,11 @@ static int p4_compute_mmcr(u64 event[], int n_ev, return -1; /* Set TTMxSEL fields. Note, units 1-3 => TTM0SEL codes 0-2 */ - mmcr1 |= (u64)(unituse[3] * 2 + unituse[2]) << MMCR1_TTM0SEL_SH; - mmcr1 |= (u64)(unituse[7] * 3 + unituse[6] * 2) << MMCR1_TTM1SEL_SH; - mmcr1 |= (u64)unituse[9] << MMCR1_TTM2SEL_SH; + mmcr1 |= (unsigned long)(unituse[3] * 2 + unituse[2]) + << MMCR1_TTM0SEL_SH; + mmcr1 |= (unsigned long)(unituse[7] * 3 + unituse[6] * 2) + << MMCR1_TTM1SEL_SH; + mmcr1 |= (unsigned long)unituse[9] << MMCR1_TTM2SEL_SH; /* Set TTCxSEL fields. */ if (unitlower & 0xe) @@ -456,7 +459,8 @@ static int p4_compute_mmcr(u64 event[], int n_ev, ttm = unit - 1; /* 2->1, 3->2 */ else ttm = unit >> 2; - mmcr1 |= (u64)ttm << (MMCR1_TD_CP_DBG0SEL_SH - 2*byte); + mmcr1 |= (unsigned long)ttm + << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte); } } @@ -519,7 +523,7 @@ static int p4_compute_mmcr(u64 event[], int n_ev, return 0; } -static void p4_disable_pmc(unsigned int pmc, u64 mmcr[]) +static void p4_disable_pmc(unsigned int pmc, unsigned long mmcr[]) { /* * Setting the PMCxSEL field to 0 disables PMC x. @@ -584,15 +588,15 @@ static int power4_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }; struct power_pmu power4_pmu = { - .n_counter = 8, - .max_alternatives = 5, - .add_fields = 0x0000001100005555ull, - .test_adder = 0x0011083300000000ull, - .compute_mmcr = p4_compute_mmcr, - .get_constraint = p4_get_constraint, - .get_alternatives = p4_get_alternatives, - .disable_pmc = p4_disable_pmc, - .n_generic = ARRAY_SIZE(p4_generic_events), - .generic_events = p4_generic_events, - .cache_events = &power4_cache_events, + .n_counter = 8, + .max_alternatives = 5, + .add_fields = 0x0000001100005555ul, + .test_adder = 0x0011083300000000ul, + .compute_mmcr = p4_compute_mmcr, + .get_constraint = p4_get_constraint, + .get_alternatives = p4_get_alternatives, + .disable_pmc = p4_disable_pmc, + .n_generic = ARRAY_SIZE(p4_generic_events), + .generic_events = p4_generic_events, + .cache_events = &power4_cache_events, }; diff --git a/arch/powerpc/kernel/power5+-pmu.c b/arch/powerpc/kernel/power5+-pmu.c index 41e5d2d958d4..aef144d503b0 100644 --- a/arch/powerpc/kernel/power5+-pmu.c +++ b/arch/powerpc/kernel/power5+-pmu.c @@ -126,20 +126,21 @@ static const int grsel_shift[8] = { }; /* Masks and values for using events from the various units */ -static u64 unit_cons[PM_LASTUNIT+1][2] = { - [PM_FPU] = { 0x3200000000ull, 0x0100000000ull }, - [PM_ISU0] = { 0x0200000000ull, 0x0080000000ull }, - [PM_ISU1] = { 0x3200000000ull, 0x3100000000ull }, - [PM_IFU] = { 0x3200000000ull, 0x2100000000ull }, - [PM_IDU] = { 0x0e00000000ull, 0x0040000000ull }, - [PM_GRS] = { 0x0e00000000ull, 0x0c40000000ull }, +static unsigned long unit_cons[PM_LASTUNIT+1][2] = { + [PM_FPU] = { 0x3200000000ul, 0x0100000000ul }, + [PM_ISU0] = { 0x0200000000ul, 0x0080000000ul }, + [PM_ISU1] = { 0x3200000000ul, 0x3100000000ul }, + [PM_IFU] = { 0x3200000000ul, 0x2100000000ul }, + [PM_IDU] = { 0x0e00000000ul, 0x0040000000ul }, + [PM_GRS] = { 0x0e00000000ul, 0x0c40000000ul }, }; -static int power5p_get_constraint(u64 event, u64 *maskp, u64 *valp) +static int power5p_get_constraint(u64 event, unsigned long *maskp, + unsigned long *valp) { int pmc, byte, unit, sh; int bit, fmask; - u64 mask = 0, value = 0; + unsigned long mask = 0, value = 0; pmc = (event >> PM_PMC_SH) & PM_PMC_MSK; if (pmc) { @@ -171,17 +172,18 @@ static int power5p_get_constraint(u64 event, u64 *maskp, u64 *valp) bit = event & 7; fmask = (bit == 6)? 7: 3; sh = grsel_shift[bit]; - mask |= (u64)fmask << sh; - value |= (u64)((event >> PM_GRS_SH) & fmask) << sh; + mask |= (unsigned long)fmask << sh; + value |= (unsigned long)((event >> PM_GRS_SH) & fmask) + << sh; } /* Set byte lane select field */ - mask |= 0xfULL << (24 - 4 * byte); - value |= (u64)unit << (24 - 4 * byte); + mask |= 0xfUL << (24 - 4 * byte); + value |= (unsigned long)unit << (24 - 4 * byte); } if (pmc < 5) { /* need a counter from PMC1-4 set */ - mask |= 0x8000000000000ull; - value |= 0x1000000000000ull; + mask |= 0x8000000000000ul; + value |= 0x1000000000000ul; } *maskp = mask; *valp = value; @@ -452,10 +454,10 @@ static int power5p_marked_instr_event(u64 event) } static int power5p_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], u64 mmcr[]) + unsigned int hwc[], unsigned long mmcr[]) { - u64 mmcr1 = 0; - u64 mmcra = 0; + unsigned long mmcr1 = 0; + unsigned long mmcra = 0; unsigned int pmc, unit, byte, psel; unsigned int ttm; int i, isbus, bit, grsel; @@ -517,7 +519,7 @@ static int power5p_compute_mmcr(u64 event[], int n_ev, continue; if (ttmuse++) return -1; - mmcr1 |= (u64)i << MMCR1_TTM0SEL_SH; + mmcr1 |= (unsigned long)i << MMCR1_TTM0SEL_SH; } ttmuse = 0; for (; i <= PM_GRS; ++i) { @@ -525,7 +527,7 @@ static int power5p_compute_mmcr(u64 event[], int n_ev, continue; if (ttmuse++) return -1; - mmcr1 |= (u64)(i & 3) << MMCR1_TTM1SEL_SH; + mmcr1 |= (unsigned long)(i & 3) << MMCR1_TTM1SEL_SH; } if (ttmuse > 1) return -1; @@ -540,10 +542,11 @@ static int power5p_compute_mmcr(u64 event[], int n_ev, unit = PM_ISU0_ALT; } else if (unit == PM_LSU1 + 1) { /* select lower word of LSU1 for this byte */ - mmcr1 |= 1ull << (MMCR1_TTM3SEL_SH + 3 - byte); + mmcr1 |= 1ul << (MMCR1_TTM3SEL_SH + 3 - byte); } ttm = unit >> 2; - mmcr1 |= (u64)ttm << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte); + mmcr1 |= (unsigned long)ttm + << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte); } /* Second pass: assign PMCs, set PMCxSEL and PMCx_ADDER_SEL fields */ @@ -568,7 +571,7 @@ static int power5p_compute_mmcr(u64 event[], int n_ev, if (isbus && (byte & 2) && (psel == 8 || psel == 0x10 || psel == 0x28)) /* add events on higher-numbered bus */ - mmcr1 |= 1ull << (MMCR1_PMC1_ADDER_SEL_SH - pmc); + mmcr1 |= 1ul << (MMCR1_PMC1_ADDER_SEL_SH - pmc); } else { /* Instructions or run cycles on PMC5/6 */ --pmc; @@ -576,7 +579,7 @@ static int power5p_compute_mmcr(u64 event[], int n_ev, if (isbus && unit == PM_GRS) { bit = psel & 7; grsel = (event[i] >> PM_GRS_SH) & PM_GRS_MSK; - mmcr1 |= (u64)grsel << grsel_shift[bit]; + mmcr1 |= (unsigned long)grsel << grsel_shift[bit]; } if (power5p_marked_instr_event(event[i])) mmcra |= MMCRA_SAMPLE_ENABLE; @@ -599,7 +602,7 @@ static int power5p_compute_mmcr(u64 event[], int n_ev, return 0; } -static void power5p_disable_pmc(unsigned int pmc, u64 mmcr[]) +static void power5p_disable_pmc(unsigned int pmc, unsigned long mmcr[]) { if (pmc <= 3) mmcr[1] &= ~(0x7fUL << MMCR1_PMCSEL_SH(pmc)); @@ -655,17 +658,17 @@ static int power5p_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }; struct power_pmu power5p_pmu = { - .n_counter = 6, - .max_alternatives = MAX_ALT, - .add_fields = 0x7000000000055ull, - .test_adder = 0x3000040000000ull, - .compute_mmcr = power5p_compute_mmcr, - .get_constraint = power5p_get_constraint, - .get_alternatives = power5p_get_alternatives, - .disable_pmc = power5p_disable_pmc, - .limited_pmc_event = power5p_limited_pmc_event, - .flags = PPMU_LIMITED_PMC5_6, - .n_generic = ARRAY_SIZE(power5p_generic_events), - .generic_events = power5p_generic_events, - .cache_events = &power5p_cache_events, + .n_counter = 6, + .max_alternatives = MAX_ALT, + .add_fields = 0x7000000000055ul, + .test_adder = 0x3000040000000ul, + .compute_mmcr = power5p_compute_mmcr, + .get_constraint = power5p_get_constraint, + .get_alternatives = power5p_get_alternatives, + .disable_pmc = power5p_disable_pmc, + .limited_pmc_event = power5p_limited_pmc_event, + .flags = PPMU_LIMITED_PMC5_6, + .n_generic = ARRAY_SIZE(power5p_generic_events), + .generic_events = power5p_generic_events, + .cache_events = &power5p_cache_events, }; diff --git a/arch/powerpc/kernel/power5-pmu.c b/arch/powerpc/kernel/power5-pmu.c index 05600b66221a..8694c73bfb52 100644 --- a/arch/powerpc/kernel/power5-pmu.c +++ b/arch/powerpc/kernel/power5-pmu.c @@ -130,20 +130,21 @@ static const int grsel_shift[8] = { }; /* Masks and values for using events from the various units */ -static u64 unit_cons[PM_LASTUNIT+1][2] = { - [PM_FPU] = { 0xc0002000000000ull, 0x00001000000000ull }, - [PM_ISU0] = { 0x00002000000000ull, 0x00000800000000ull }, - [PM_ISU1] = { 0xc0002000000000ull, 0xc0001000000000ull }, - [PM_IFU] = { 0xc0002000000000ull, 0x80001000000000ull }, - [PM_IDU] = { 0x30002000000000ull, 0x00000400000000ull }, - [PM_GRS] = { 0x30002000000000ull, 0x30000400000000ull }, +static unsigned long unit_cons[PM_LASTUNIT+1][2] = { + [PM_FPU] = { 0xc0002000000000ul, 0x00001000000000ul }, + [PM_ISU0] = { 0x00002000000000ul, 0x00000800000000ul }, + [PM_ISU1] = { 0xc0002000000000ul, 0xc0001000000000ul }, + [PM_IFU] = { 0xc0002000000000ul, 0x80001000000000ul }, + [PM_IDU] = { 0x30002000000000ul, 0x00000400000000ul }, + [PM_GRS] = { 0x30002000000000ul, 0x30000400000000ul }, }; -static int power5_get_constraint(u64 event, u64 *maskp, u64 *valp) +static int power5_get_constraint(u64 event, unsigned long *maskp, + unsigned long *valp) { int pmc, byte, unit, sh; int bit, fmask; - u64 mask = 0, value = 0; + unsigned long mask = 0, value = 0; int grp = -1; pmc = (event >> PM_PMC_SH) & PM_PMC_MSK; @@ -178,8 +179,9 @@ static int power5_get_constraint(u64 event, u64 *maskp, u64 *valp) bit = event & 7; fmask = (bit == 6)? 7: 3; sh = grsel_shift[bit]; - mask |= (u64)fmask << sh; - value |= (u64)((event >> PM_GRS_SH) & fmask) << sh; + mask |= (unsigned long)fmask << sh; + value |= (unsigned long)((event >> PM_GRS_SH) & fmask) + << sh; } /* * Bus events on bytes 0 and 2 can be counted @@ -188,22 +190,22 @@ static int power5_get_constraint(u64 event, u64 *maskp, u64 *valp) if (!pmc) grp = byte & 1; /* Set byte lane select field */ - mask |= 0xfULL << (24 - 4 * byte); - value |= (u64)unit << (24 - 4 * byte); + mask |= 0xfUL << (24 - 4 * byte); + value |= (unsigned long)unit << (24 - 4 * byte); } if (grp == 0) { /* increment PMC1/2 field */ - mask |= 0x200000000ull; - value |= 0x080000000ull; + mask |= 0x200000000ul; + value |= 0x080000000ul; } else if (grp == 1) { /* increment PMC3/4 field */ - mask |= 0x40000000ull; - value |= 0x10000000ull; + mask |= 0x40000000ul; + value |= 0x10000000ul; } if (pmc < 5) { /* need a counter from PMC1-4 set */ - mask |= 0x8000000000000ull; - value |= 0x1000000000000ull; + mask |= 0x8000000000000ul; + value |= 0x1000000000000ul; } *maskp = mask; *valp = value; @@ -383,10 +385,10 @@ static int power5_marked_instr_event(u64 event) } static int power5_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], u64 mmcr[]) + unsigned int hwc[], unsigned long mmcr[]) { - u64 mmcr1 = 0; - u64 mmcra = 0; + unsigned long mmcr1 = 0; + unsigned long mmcra = 0; unsigned int pmc, unit, byte, psel; unsigned int ttm, grp; int i, isbus, bit, grsel; @@ -457,7 +459,7 @@ static int power5_compute_mmcr(u64 event[], int n_ev, continue; if (ttmuse++) return -1; - mmcr1 |= (u64)i << MMCR1_TTM0SEL_SH; + mmcr1 |= (unsigned long)i << MMCR1_TTM0SEL_SH; } ttmuse = 0; for (; i <= PM_GRS; ++i) { @@ -465,7 +467,7 @@ static int power5_compute_mmcr(u64 event[], int n_ev, continue; if (ttmuse++) return -1; - mmcr1 |= (u64)(i & 3) << MMCR1_TTM1SEL_SH; + mmcr1 |= (unsigned long)(i & 3) << MMCR1_TTM1SEL_SH; } if (ttmuse > 1) return -1; @@ -480,10 +482,11 @@ static int power5_compute_mmcr(u64 event[], int n_ev, unit = PM_ISU0_ALT; } else if (unit == PM_LSU1 + 1) { /* select lower word of LSU1 for this byte */ - mmcr1 |= 1ull << (MMCR1_TTM3SEL_SH + 3 - byte); + mmcr1 |= 1ul << (MMCR1_TTM3SEL_SH + 3 - byte); } ttm = unit >> 2; - mmcr1 |= (u64)ttm << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte); + mmcr1 |= (unsigned long)ttm + << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte); } /* Second pass: assign PMCs, set PMCxSEL and PMCx_ADDER_SEL fields */ @@ -513,7 +516,7 @@ static int power5_compute_mmcr(u64 event[], int n_ev, --pmc; if ((psel == 8 || psel == 0x10) && isbus && (byte & 2)) /* add events on higher-numbered bus */ - mmcr1 |= 1ull << (MMCR1_PMC1_ADDER_SEL_SH - pmc); + mmcr1 |= 1ul << (MMCR1_PMC1_ADDER_SEL_SH - pmc); } else { /* Instructions or run cycles on PMC5/6 */ --pmc; @@ -521,7 +524,7 @@ static int power5_compute_mmcr(u64 event[], int n_ev, if (isbus && unit == PM_GRS) { bit = psel & 7; grsel = (event[i] >> PM_GRS_SH) & PM_GRS_MSK; - mmcr1 |= (u64)grsel << grsel_shift[bit]; + mmcr1 |= (unsigned long)grsel << grsel_shift[bit]; } if (power5_marked_instr_event(event[i])) mmcra |= MMCRA_SAMPLE_ENABLE; @@ -541,7 +544,7 @@ static int power5_compute_mmcr(u64 event[], int n_ev, return 0; } -static void power5_disable_pmc(unsigned int pmc, u64 mmcr[]) +static void power5_disable_pmc(unsigned int pmc, unsigned long mmcr[]) { if (pmc <= 3) mmcr[1] &= ~(0x7fUL << MMCR1_PMCSEL_SH(pmc)); @@ -597,15 +600,15 @@ static int power5_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }; struct power_pmu power5_pmu = { - .n_counter = 6, - .max_alternatives = MAX_ALT, - .add_fields = 0x7000090000555ull, - .test_adder = 0x3000490000000ull, - .compute_mmcr = power5_compute_mmcr, - .get_constraint = power5_get_constraint, - .get_alternatives = power5_get_alternatives, - .disable_pmc = power5_disable_pmc, - .n_generic = ARRAY_SIZE(power5_generic_events), - .generic_events = power5_generic_events, - .cache_events = &power5_cache_events, + .n_counter = 6, + .max_alternatives = MAX_ALT, + .add_fields = 0x7000090000555ul, + .test_adder = 0x3000490000000ul, + .compute_mmcr = power5_compute_mmcr, + .get_constraint = power5_get_constraint, + .get_alternatives = power5_get_alternatives, + .disable_pmc = power5_disable_pmc, + .n_generic = ARRAY_SIZE(power5_generic_events), + .generic_events = power5_generic_events, + .cache_events = &power5_cache_events, }; diff --git a/arch/powerpc/kernel/power6-pmu.c b/arch/powerpc/kernel/power6-pmu.c index 46f74bebcfd9..8898622ac28c 100644 --- a/arch/powerpc/kernel/power6-pmu.c +++ b/arch/powerpc/kernel/power6-pmu.c @@ -41,9 +41,9 @@ #define MMCR1_NESTSEL_SH 45 #define MMCR1_NESTSEL_MSK 0x7 #define MMCR1_NESTSEL(m) (((m) >> MMCR1_NESTSEL_SH) & MMCR1_NESTSEL_MSK) -#define MMCR1_PMC1_LLA ((u64)1 << 44) -#define MMCR1_PMC1_LLA_VALUE ((u64)1 << 39) -#define MMCR1_PMC1_ADDR_SEL ((u64)1 << 35) +#define MMCR1_PMC1_LLA (1ul << 44) +#define MMCR1_PMC1_LLA_VALUE (1ul << 39) +#define MMCR1_PMC1_ADDR_SEL (1ul << 35) #define MMCR1_PMC1SEL_SH 24 #define MMCR1_PMCSEL_SH(n) (MMCR1_PMC1SEL_SH - (n) * 8) #define MMCR1_PMCSEL_MSK 0xff @@ -173,10 +173,10 @@ static int power6_marked_instr_event(u64 event) * Assign PMC numbers and compute MMCR1 value for a set of events */ static int p6_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], u64 mmcr[]) + unsigned int hwc[], unsigned long mmcr[]) { - u64 mmcr1 = 0; - u64 mmcra = 0; + unsigned long mmcr1 = 0; + unsigned long mmcra = 0; int i; unsigned int pmc, ev, b, u, s, psel; unsigned int ttmset = 0; @@ -215,7 +215,7 @@ static int p6_compute_mmcr(u64 event[], int n_ev, /* check for conflict on this byte of event bus */ if ((ttmset & (1 << b)) && MMCR1_TTMSEL(mmcr1, b) != u) return -1; - mmcr1 |= (u64)u << MMCR1_TTMSEL_SH(b); + mmcr1 |= (unsigned long)u << MMCR1_TTMSEL_SH(b); ttmset |= 1 << b; if (u == 5) { /* Nest events have a further mux */ @@ -224,7 +224,7 @@ static int p6_compute_mmcr(u64 event[], int n_ev, MMCR1_NESTSEL(mmcr1) != s) return -1; ttmset |= 0x10; - mmcr1 |= (u64)s << MMCR1_NESTSEL_SH; + mmcr1 |= (unsigned long)s << MMCR1_NESTSEL_SH; } if (0x30 <= psel && psel <= 0x3d) { /* these need the PMCx_ADDR_SEL bits */ @@ -243,7 +243,7 @@ static int p6_compute_mmcr(u64 event[], int n_ev, if (power6_marked_instr_event(event[i])) mmcra |= MMCRA_SAMPLE_ENABLE; if (pmc < 4) - mmcr1 |= (u64)psel << MMCR1_PMCSEL_SH(pmc); + mmcr1 |= (unsigned long)psel << MMCR1_PMCSEL_SH(pmc); } mmcr[0] = 0; if (pmc_inuse & 1) @@ -265,10 +265,11 @@ static int p6_compute_mmcr(u64 event[], int n_ev, * 20-23, 24-27, 28-31 ditto for bytes 1, 2, 3 * 32-34 select field: nest (subunit) event selector */ -static int p6_get_constraint(u64 event, u64 *maskp, u64 *valp) +static int p6_get_constraint(u64 event, unsigned long *maskp, + unsigned long *valp) { int pmc, byte, sh, subunit; - u64 mask = 0, value = 0; + unsigned long mask = 0, value = 0; pmc = (event >> PM_PMC_SH) & PM_PMC_MSK; if (pmc) { @@ -282,11 +283,11 @@ static int p6_get_constraint(u64 event, u64 *maskp, u64 *valp) byte = (event >> PM_BYTE_SH) & PM_BYTE_MSK; sh = byte * 4 + (16 - PM_UNIT_SH); mask |= PM_UNIT_MSKS << sh; - value |= (u64)(event & PM_UNIT_MSKS) << sh; + value |= (unsigned long)(event & PM_UNIT_MSKS) << sh; if ((event & PM_UNIT_MSKS) == (5 << PM_UNIT_SH)) { subunit = (event >> PM_SUBUNIT_SH) & PM_SUBUNIT_MSK; - mask |= (u64)PM_SUBUNIT_MSK << 32; - value |= (u64)subunit << 32; + mask |= (unsigned long)PM_SUBUNIT_MSK << 32; + value |= (unsigned long)subunit << 32; } } if (pmc <= 4) { @@ -458,7 +459,7 @@ static int p6_get_alternatives(u64 event, unsigned int flags, u64 alt[]) return nalt; } -static void p6_disable_pmc(unsigned int pmc, u64 mmcr[]) +static void p6_disable_pmc(unsigned int pmc, unsigned long mmcr[]) { /* Set PMCxSEL to 0 to disable PMCx */ if (pmc <= 3) @@ -516,17 +517,17 @@ static int power6_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }; struct power_pmu power6_pmu = { - .n_counter = 6, - .max_alternatives = MAX_ALT, - .add_fields = 0x1555, - .test_adder = 0x3000, - .compute_mmcr = p6_compute_mmcr, - .get_constraint = p6_get_constraint, - .get_alternatives = p6_get_alternatives, - .disable_pmc = p6_disable_pmc, - .limited_pmc_event = p6_limited_pmc_event, - .flags = PPMU_LIMITED_PMC5_6 | PPMU_ALT_SIPR, - .n_generic = ARRAY_SIZE(power6_generic_events), - .generic_events = power6_generic_events, - .cache_events = &power6_cache_events, + .n_counter = 6, + .max_alternatives = MAX_ALT, + .add_fields = 0x1555, + .test_adder = 0x3000, + .compute_mmcr = p6_compute_mmcr, + .get_constraint = p6_get_constraint, + .get_alternatives = p6_get_alternatives, + .disable_pmc = p6_disable_pmc, + .limited_pmc_event = p6_limited_pmc_event, + .flags = PPMU_LIMITED_PMC5_6 | PPMU_ALT_SIPR, + .n_generic = ARRAY_SIZE(power6_generic_events), + .generic_events = power6_generic_events, + .cache_events = &power6_cache_events, }; diff --git a/arch/powerpc/kernel/power7-pmu.c b/arch/powerpc/kernel/power7-pmu.c index b72e7a19d054..658d1ae436a0 100644 --- a/arch/powerpc/kernel/power7-pmu.c +++ b/arch/powerpc/kernel/power7-pmu.c @@ -71,10 +71,11 @@ * 0-9: Count of events needing PMC1..PMC5 */ -static int power7_get_constraint(u64 event, u64 *maskp, u64 *valp) +static int power7_get_constraint(u64 event, unsigned long *maskp, + unsigned long *valp) { int pmc, sh; - u64 mask = 0, value = 0; + unsigned long mask = 0, value = 0; pmc = (event >> PM_PMC_SH) & PM_PMC_MSK; if (pmc) { @@ -224,10 +225,10 @@ static int power7_marked_instr_event(u64 event) } static int power7_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], u64 mmcr[]) + unsigned int hwc[], unsigned long mmcr[]) { - u64 mmcr1 = 0; - u64 mmcra = 0; + unsigned long mmcr1 = 0; + unsigned long mmcra = 0; unsigned int pmc, unit, combine, l2sel, psel; unsigned int pmc_inuse = 0; int i; @@ -265,11 +266,14 @@ static int power7_compute_mmcr(u64 event[], int n_ev, --pmc; } if (pmc <= 3) { - mmcr1 |= (u64) unit << (MMCR1_TTM0SEL_SH - 4 * pmc); - mmcr1 |= (u64) combine << (MMCR1_PMC1_COMBINE_SH - pmc); + mmcr1 |= (unsigned long) unit + << (MMCR1_TTM0SEL_SH - 4 * pmc); + mmcr1 |= (unsigned long) combine + << (MMCR1_PMC1_COMBINE_SH - pmc); mmcr1 |= psel << MMCR1_PMCSEL_SH(pmc); if (unit == 6) /* L2 events */ - mmcr1 |= (u64) l2sel << MMCR1_L2SEL_SH; + mmcr1 |= (unsigned long) l2sel + << MMCR1_L2SEL_SH; } if (power7_marked_instr_event(event[i])) mmcra |= MMCRA_SAMPLE_ENABLE; @@ -287,10 +291,10 @@ static int power7_compute_mmcr(u64 event[], int n_ev, return 0; } -static void power7_disable_pmc(unsigned int pmc, u64 mmcr[]) +static void power7_disable_pmc(unsigned int pmc, unsigned long mmcr[]) { if (pmc <= 3) - mmcr[1] &= ~(0xffULL << MMCR1_PMCSEL_SH(pmc)); + mmcr[1] &= ~(0xffUL << MMCR1_PMCSEL_SH(pmc)); } static int power7_generic_events[] = { @@ -343,15 +347,15 @@ static int power7_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }; struct power_pmu power7_pmu = { - .n_counter = 6, - .max_alternatives = MAX_ALT + 1, - .add_fields = 0x1555ull, - .test_adder = 0x3000ull, - .compute_mmcr = power7_compute_mmcr, - .get_constraint = power7_get_constraint, - .get_alternatives = power7_get_alternatives, - .disable_pmc = power7_disable_pmc, - .n_generic = ARRAY_SIZE(power7_generic_events), - .generic_events = power7_generic_events, - .cache_events = &power7_cache_events, + .n_counter = 6, + .max_alternatives = MAX_ALT + 1, + .add_fields = 0x1555ul, + .test_adder = 0x3000ul, + .compute_mmcr = power7_compute_mmcr, + .get_constraint = power7_get_constraint, + .get_alternatives = power7_get_alternatives, + .disable_pmc = power7_disable_pmc, + .n_generic = ARRAY_SIZE(power7_generic_events), + .generic_events = power7_generic_events, + .cache_events = &power7_cache_events, }; diff --git a/arch/powerpc/kernel/ppc970-pmu.c b/arch/powerpc/kernel/ppc970-pmu.c index ba0a357a89f4..3ed88333412f 100644 --- a/arch/powerpc/kernel/ppc970-pmu.c +++ b/arch/powerpc/kernel/ppc970-pmu.c @@ -183,7 +183,7 @@ static int p970_marked_instr_event(u64 event) } /* Masks and values for using events from the various units */ -static u64 unit_cons[PM_LASTUNIT+1][2] = { +static unsigned long unit_cons[PM_LASTUNIT+1][2] = { [PM_FPU] = { 0xc80000000000ull, 0x040000000000ull }, [PM_VPU] = { 0xc80000000000ull, 0xc40000000000ull }, [PM_ISU] = { 0x080000000000ull, 0x020000000000ull }, @@ -192,10 +192,11 @@ static u64 unit_cons[PM_LASTUNIT+1][2] = { [PM_STS] = { 0x380000000000ull, 0x310000000000ull }, }; -static int p970_get_constraint(u64 event, u64 *maskp, u64 *valp) +static int p970_get_constraint(u64 event, unsigned long *maskp, + unsigned long *valp) { int pmc, byte, unit, sh, spcsel; - u64 mask = 0, value = 0; + unsigned long mask = 0, value = 0; int grp = -1; pmc = (event >> PM_PMC_SH) & PM_PMC_MSK; @@ -222,7 +223,7 @@ static int p970_get_constraint(u64 event, u64 *maskp, u64 *valp) grp = byte & 1; /* Set byte lane select field */ mask |= 0xfULL << (28 - 4 * byte); - value |= (u64)unit << (28 - 4 * byte); + value |= (unsigned long)unit << (28 - 4 * byte); } if (grp == 0) { /* increment PMC1/2/5/6 field */ @@ -236,7 +237,7 @@ static int p970_get_constraint(u64 event, u64 *maskp, u64 *valp) spcsel = (event >> PM_SPCSEL_SH) & PM_SPCSEL_MSK; if (spcsel) { mask |= 3ull << 48; - value |= (u64)spcsel << 48; + value |= (unsigned long)spcsel << 48; } *maskp = mask; *valp = value; @@ -257,9 +258,9 @@ static int p970_get_alternatives(u64 event, unsigned int flags, u64 alt[]) } static int p970_compute_mmcr(u64 event[], int n_ev, - unsigned int hwc[], u64 mmcr[]) + unsigned int hwc[], unsigned long mmcr[]) { - u64 mmcr0 = 0, mmcr1 = 0, mmcra = 0; + unsigned long mmcr0 = 0, mmcr1 = 0, mmcra = 0; unsigned int pmc, unit, byte, psel; unsigned int ttm, grp; unsigned int pmc_inuse = 0; @@ -320,7 +321,7 @@ static int p970_compute_mmcr(u64 event[], int n_ev, continue; ttm = unitmap[i]; ++ttmuse[(ttm >> 2) & 1]; - mmcr1 |= (u64)(ttm & ~4) << MMCR1_TTM1SEL_SH; + mmcr1 |= (unsigned long)(ttm & ~4) << MMCR1_TTM1SEL_SH; } /* Check only one unit per TTMx */ if (ttmuse[0] > 1 || ttmuse[1] > 1) @@ -340,7 +341,8 @@ static int p970_compute_mmcr(u64 event[], int n_ev, if (unit == PM_LSU1L && byte >= 2) mmcr1 |= 1ull << (MMCR1_TTM3SEL_SH + 3 - byte); } - mmcr1 |= (u64)ttm << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte); + mmcr1 |= (unsigned long)ttm + << (MMCR1_TD_CP_DBG0SEL_SH - 2 * byte); } /* Second pass: assign PMCs, set PMCxSEL and PMCx_ADDER_SEL fields */ @@ -386,7 +388,8 @@ static int p970_compute_mmcr(u64 event[], int n_ev, for (pmc = 0; pmc < 2; ++pmc) mmcr0 |= pmcsel[pmc] << (MMCR0_PMC1SEL_SH - 7 * pmc); for (; pmc < 8; ++pmc) - mmcr1 |= (u64)pmcsel[pmc] << (MMCR1_PMC3SEL_SH - 5 * (pmc - 2)); + mmcr1 |= (unsigned long)pmcsel[pmc] + << (MMCR1_PMC3SEL_SH - 5 * (pmc - 2)); if (pmc_inuse & 1) mmcr0 |= MMCR0_PMC1CE; if (pmc_inuse & 0xfe) @@ -401,7 +404,7 @@ static int p970_compute_mmcr(u64 event[], int n_ev, return 0; } -static void p970_disable_pmc(unsigned int pmc, u64 mmcr[]) +static void p970_disable_pmc(unsigned int pmc, unsigned long mmcr[]) { int shift, i; @@ -468,15 +471,15 @@ static int ppc970_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }; struct power_pmu ppc970_pmu = { - .n_counter = 8, - .max_alternatives = 2, - .add_fields = 0x001100005555ull, - .test_adder = 0x013300000000ull, - .compute_mmcr = p970_compute_mmcr, - .get_constraint = p970_get_constraint, - .get_alternatives = p970_get_alternatives, - .disable_pmc = p970_disable_pmc, - .n_generic = ARRAY_SIZE(ppc970_generic_events), - .generic_events = ppc970_generic_events, - .cache_events = &ppc970_cache_events, + .n_counter = 8, + .max_alternatives = 2, + .add_fields = 0x001100005555ull, + .test_adder = 0x013300000000ull, + .compute_mmcr = p970_compute_mmcr, + .get_constraint = p970_get_constraint, + .get_alternatives = p970_get_alternatives, + .disable_pmc = p970_disable_pmc, + .n_generic = ARRAY_SIZE(ppc970_generic_events), + .generic_events = ppc970_generic_events, + .cache_events = &ppc970_cache_events, }; -- cgit v1.2.3 From 079b3c569c87819e7a19d9b9f51d4746fc47bf9a Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Wed, 17 Jun 2009 21:52:09 +1000 Subject: perf_counter: powerpc: Change how processor-specific back-ends get selected At present, the powerpc generic (processor-independent) perf_counter code has list of processor back-end modules, and at initialization, it looks at the PVR (processor version register) and has a switch statement to select a suitable processor-specific back-end. This is going to become inconvenient as we add more processor-specific back-ends, so this inverts the order: now each back-end checks whether it applies to the current processor, and registers itself if so. Furthermore, instead of looking at the PVR, back-ends now check the cur_cpu_spec->oprofile_cpu_type string and match on that. Lastly, each back-end now specifies a name for itself so the core can print a nice message when a back-end registers itself. This doesn't provide any support for unregistering back-ends, but that wouldn't be hard to do and would allow back-ends to be modules. Signed-off-by: Paul Mackerras Cc: Peter Zijlstra Cc: linuxppc-dev@ozlabs.org Cc: benh@kernel.crashing.org LKML-Reference: <19000.55529.762227.518531@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- arch/powerpc/include/asm/perf_counter.h | 5 ++-- arch/powerpc/kernel/perf_counter.c | 44 ++++++--------------------------- arch/powerpc/kernel/power4-pmu.c | 15 ++++++++++- arch/powerpc/kernel/power5+-pmu.c | 16 +++++++++++- arch/powerpc/kernel/power5-pmu.c | 15 ++++++++++- arch/powerpc/kernel/power6-pmu.c | 15 ++++++++++- arch/powerpc/kernel/power7-pmu.c | 15 ++++++++++- arch/powerpc/kernel/ppc970-pmu.c | 16 +++++++++++- 8 files changed, 96 insertions(+), 45 deletions(-) (limited to 'arch') diff --git a/arch/powerpc/include/asm/perf_counter.h b/arch/powerpc/include/asm/perf_counter.h index 2ceb0fefa93a..8ccd4e155768 100644 --- a/arch/powerpc/include/asm/perf_counter.h +++ b/arch/powerpc/include/asm/perf_counter.h @@ -21,6 +21,7 @@ * describe the PMU on a particular POWER-family CPU. */ struct power_pmu { + const char *name; int n_counter; int max_alternatives; unsigned long add_fields; @@ -41,8 +42,6 @@ struct power_pmu { [PERF_COUNT_HW_CACHE_RESULT_MAX]; }; -extern struct power_pmu *ppmu; - /* * Values for power_pmu.flags */ @@ -56,6 +55,8 @@ extern struct power_pmu *ppmu; #define PPMU_LIMITED_PMC_REQD 2 /* have to put this on a limited PMC */ #define PPMU_ONLY_COUNT_RUN 4 /* only counting in run state */ +extern int register_power_pmu(struct power_pmu *); + struct pt_regs; extern unsigned long perf_misc_flags(struct pt_regs *regs); extern unsigned long perf_instruction_pointer(struct pt_regs *regs); diff --git a/arch/powerpc/kernel/perf_counter.c b/arch/powerpc/kernel/perf_counter.c index 9300638b8c26..25e656c14945 100644 --- a/arch/powerpc/kernel/perf_counter.c +++ b/arch/powerpc/kernel/perf_counter.c @@ -1214,42 +1214,14 @@ void hw_perf_counter_setup(int cpu) cpuhw->mmcr[0] = MMCR0_FC; } -extern struct power_pmu power4_pmu; -extern struct power_pmu ppc970_pmu; -extern struct power_pmu power5_pmu; -extern struct power_pmu power5p_pmu; -extern struct power_pmu power6_pmu; -extern struct power_pmu power7_pmu; - -static int init_perf_counters(void) +int register_power_pmu(struct power_pmu *pmu) { - unsigned long pvr; - - /* XXX should get this from cputable */ - pvr = mfspr(SPRN_PVR); - switch (PVR_VER(pvr)) { - case PV_POWER4: - case PV_POWER4p: - ppmu = &power4_pmu; - break; - case PV_970: - case PV_970FX: - case PV_970MP: - ppmu = &ppc970_pmu; - break; - case PV_POWER5: - ppmu = &power5_pmu; - break; - case PV_POWER5p: - ppmu = &power5p_pmu; - break; - case 0x3e: - ppmu = &power6_pmu; - break; - case 0x3f: - ppmu = &power7_pmu; - break; - } + if (ppmu) + return -EBUSY; /* something's already registered */ + + ppmu = pmu; + pr_info("%s performance monitor hardware support registered\n", + pmu->name); /* * Use FCHV to ignore kernel events if MSR.HV is set. @@ -1259,5 +1231,3 @@ static int init_perf_counters(void) return 0; } - -arch_initcall(init_perf_counters); diff --git a/arch/powerpc/kernel/power4-pmu.c b/arch/powerpc/kernel/power4-pmu.c index 81a1708f83b2..db90b0c5c27b 100644 --- a/arch/powerpc/kernel/power4-pmu.c +++ b/arch/powerpc/kernel/power4-pmu.c @@ -10,7 +10,9 @@ */ #include #include +#include #include +#include /* * Bits in event code for POWER4 @@ -587,7 +589,8 @@ static int power4_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }, }; -struct power_pmu power4_pmu = { +static struct power_pmu power4_pmu = { + .name = "POWER4/4+", .n_counter = 8, .max_alternatives = 5, .add_fields = 0x0000001100005555ul, @@ -600,3 +603,13 @@ struct power_pmu power4_pmu = { .generic_events = p4_generic_events, .cache_events = &power4_cache_events, }; + +static int init_power4_pmu(void) +{ + if (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power4")) + return -ENODEV; + + return register_power_pmu(&power4_pmu); +} + +arch_initcall(init_power4_pmu); diff --git a/arch/powerpc/kernel/power5+-pmu.c b/arch/powerpc/kernel/power5+-pmu.c index aef144d503b0..f4adca8e98a4 100644 --- a/arch/powerpc/kernel/power5+-pmu.c +++ b/arch/powerpc/kernel/power5+-pmu.c @@ -10,7 +10,9 @@ */ #include #include +#include #include +#include /* * Bits in event code for POWER5+ (POWER5 GS) and POWER5++ (POWER5 GS DD3) @@ -657,7 +659,8 @@ static int power5p_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }, }; -struct power_pmu power5p_pmu = { +static struct power_pmu power5p_pmu = { + .name = "POWER5+/++", .n_counter = 6, .max_alternatives = MAX_ALT, .add_fields = 0x7000000000055ul, @@ -672,3 +675,14 @@ struct power_pmu power5p_pmu = { .generic_events = power5p_generic_events, .cache_events = &power5p_cache_events, }; + +static int init_power5p_pmu(void) +{ + if (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power5+") + && strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power5++")) + return -ENODEV; + + return register_power_pmu(&power5p_pmu); +} + +arch_initcall(init_power5p_pmu); diff --git a/arch/powerpc/kernel/power5-pmu.c b/arch/powerpc/kernel/power5-pmu.c index 8694c73bfb52..29b2c6c0e83a 100644 --- a/arch/powerpc/kernel/power5-pmu.c +++ b/arch/powerpc/kernel/power5-pmu.c @@ -10,7 +10,9 @@ */ #include #include +#include #include +#include /* * Bits in event code for POWER5 (not POWER5++) @@ -599,7 +601,8 @@ static int power5_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }, }; -struct power_pmu power5_pmu = { +static struct power_pmu power5_pmu = { + .name = "POWER5", .n_counter = 6, .max_alternatives = MAX_ALT, .add_fields = 0x7000090000555ul, @@ -612,3 +615,13 @@ struct power_pmu power5_pmu = { .generic_events = power5_generic_events, .cache_events = &power5_cache_events, }; + +static int init_power5_pmu(void) +{ + if (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power5")) + return -ENODEV; + + return register_power_pmu(&power5_pmu); +} + +arch_initcall(init_power5_pmu); diff --git a/arch/powerpc/kernel/power6-pmu.c b/arch/powerpc/kernel/power6-pmu.c index 8898622ac28c..09ae5bf5bda7 100644 --- a/arch/powerpc/kernel/power6-pmu.c +++ b/arch/powerpc/kernel/power6-pmu.c @@ -10,7 +10,9 @@ */ #include #include +#include #include +#include /* * Bits in event code for POWER6 @@ -516,7 +518,8 @@ static int power6_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }, }; -struct power_pmu power6_pmu = { +static struct power_pmu power6_pmu = { + .name = "POWER6", .n_counter = 6, .max_alternatives = MAX_ALT, .add_fields = 0x1555, @@ -531,3 +534,13 @@ struct power_pmu power6_pmu = { .generic_events = power6_generic_events, .cache_events = &power6_cache_events, }; + +static int init_power6_pmu(void) +{ + if (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power6")) + return -ENODEV; + + return register_power_pmu(&power6_pmu); +} + +arch_initcall(init_power6_pmu); diff --git a/arch/powerpc/kernel/power7-pmu.c b/arch/powerpc/kernel/power7-pmu.c index 658d1ae436a0..5d755ef7ac8f 100644 --- a/arch/powerpc/kernel/power7-pmu.c +++ b/arch/powerpc/kernel/power7-pmu.c @@ -10,7 +10,9 @@ */ #include #include +#include #include +#include /* * Bits in event code for POWER7 @@ -346,7 +348,8 @@ static int power7_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }, }; -struct power_pmu power7_pmu = { +static struct power_pmu power7_pmu = { + .name = "POWER7", .n_counter = 6, .max_alternatives = MAX_ALT + 1, .add_fields = 0x1555ul, @@ -359,3 +362,13 @@ struct power_pmu power7_pmu = { .generic_events = power7_generic_events, .cache_events = &power7_cache_events, }; + +static int init_power7_pmu(void) +{ + if (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/power7")) + return -ENODEV; + + return register_power_pmu(&power7_pmu); +} + +arch_initcall(init_power7_pmu); diff --git a/arch/powerpc/kernel/ppc970-pmu.c b/arch/powerpc/kernel/ppc970-pmu.c index 3ed88333412f..6637c87fe70e 100644 --- a/arch/powerpc/kernel/ppc970-pmu.c +++ b/arch/powerpc/kernel/ppc970-pmu.c @@ -10,7 +10,9 @@ */ #include #include +#include #include +#include /* * Bits in event code for PPC970 @@ -470,7 +472,8 @@ static int ppc970_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { }, }; -struct power_pmu ppc970_pmu = { +static struct power_pmu ppc970_pmu = { + .name = "PPC970/FX/MP", .n_counter = 8, .max_alternatives = 2, .add_fields = 0x001100005555ull, @@ -483,3 +486,14 @@ struct power_pmu ppc970_pmu = { .generic_events = ppc970_generic_events, .cache_events = &ppc970_cache_events, }; + +static int init_ppc970_pmu(void) +{ + if (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/970") + && strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc64/970MP")) + return -ENODEV; + + return register_power_pmu(&ppc970_pmu); +} + +arch_initcall(init_ppc970_pmu); -- cgit v1.2.3 From 98fb1807b97e3e631b940f67544e265c64b984dc Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Wed, 17 Jun 2009 21:53:10 +1000 Subject: perf_counter: powerpc: Make powerpc perf_counter code safe for 32-bit kernels This abstracts a few things in arch/powerpc/kernel/perf_counter.c that are specific to 64-bit kernels, and provides definitions for 32-bit kernels. In particular, * Only 64-bit has MMCRA and the bits in it that give information about a PMU interrupt (sampled PR, HV, slot number etc.) * Only 64-bit has the lppaca and the lppaca->pmcregs_in_use field * Use of SDAR is confined to 64-bit for now * Only 64-bit has soft/lazy interrupt disable and therefore pseudo-NMIs (interrupts that occur while interrupts are soft-disabled) * Only 64-bit has PMC7 and PMC8 * Only 64-bit has the MSR_HV bit. This also fixes the types used in a couple of places, where we were using long types for things that need to be 64-bit. Signed-off-by: Paul Mackerras Cc: Peter Zijlstra Cc: linuxppc-dev@ozlabs.org Cc: benh@kernel.crashing.org LKML-Reference: <19000.55590.634126.876084@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- arch/powerpc/kernel/perf_counter.c | 193 +++++++++++++++++++++++++------------ 1 file changed, 133 insertions(+), 60 deletions(-) (limited to 'arch') diff --git a/arch/powerpc/kernel/perf_counter.c b/arch/powerpc/kernel/perf_counter.c index 25e656c14945..809fdf94b95f 100644 --- a/arch/powerpc/kernel/perf_counter.c +++ b/arch/powerpc/kernel/perf_counter.c @@ -46,6 +46,115 @@ struct power_pmu *ppmu; */ static unsigned int freeze_counters_kernel = MMCR0_FCS; +/* + * 32-bit doesn't have MMCRA but does have an MMCR2, + * and a few other names are different. + */ +#ifdef CONFIG_PPC32 + +#define MMCR0_FCHV 0 +#define MMCR0_PMCjCE MMCR0_PMCnCE + +#define SPRN_MMCRA SPRN_MMCR2 +#define MMCRA_SAMPLE_ENABLE 0 + +static inline unsigned long perf_ip_adjust(struct pt_regs *regs) +{ + return 0; +} +static inline void perf_set_pmu_inuse(int inuse) { } +static inline void perf_get_data_addr(struct pt_regs *regs, u64 *addrp) { } +static inline u32 perf_get_misc_flags(struct pt_regs *regs) +{ + return 0; +} +static inline void perf_read_regs(struct pt_regs *regs) { } +static inline int perf_intr_is_nmi(struct pt_regs *regs) +{ + return 0; +} + +#endif /* CONFIG_PPC32 */ + +/* + * Things that are specific to 64-bit implementations. + */ +#ifdef CONFIG_PPC64 + +static inline unsigned long perf_ip_adjust(struct pt_regs *regs) +{ + unsigned long mmcra = regs->dsisr; + + if ((mmcra & MMCRA_SAMPLE_ENABLE) && !(ppmu->flags & PPMU_ALT_SIPR)) { + unsigned long slot = (mmcra & MMCRA_SLOT) >> MMCRA_SLOT_SHIFT; + if (slot > 1) + return 4 * (slot - 1); + } + return 0; +} + +static inline void perf_set_pmu_inuse(int inuse) +{ + get_lppaca()->pmcregs_in_use = inuse; +} + +/* + * The user wants a data address recorded. + * If we're not doing instruction sampling, give them the SDAR + * (sampled data address). If we are doing instruction sampling, then + * only give them the SDAR if it corresponds to the instruction + * pointed to by SIAR; this is indicated by the [POWER6_]MMCRA_SDSYNC + * bit in MMCRA. + */ +static inline void perf_get_data_addr(struct pt_regs *regs, u64 *addrp) +{ + unsigned long mmcra = regs->dsisr; + unsigned long sdsync = (ppmu->flags & PPMU_ALT_SIPR) ? + POWER6_MMCRA_SDSYNC : MMCRA_SDSYNC; + + if (!(mmcra & MMCRA_SAMPLE_ENABLE) || (mmcra & sdsync)) + *addrp = mfspr(SPRN_SDAR); +} + +static inline u32 perf_get_misc_flags(struct pt_regs *regs) +{ + unsigned long mmcra = regs->dsisr; + + if (TRAP(regs) != 0xf00) + return 0; /* not a PMU interrupt */ + + if (ppmu->flags & PPMU_ALT_SIPR) { + if (mmcra & POWER6_MMCRA_SIHV) + return PERF_EVENT_MISC_HYPERVISOR; + return (mmcra & POWER6_MMCRA_SIPR) ? + PERF_EVENT_MISC_USER : PERF_EVENT_MISC_KERNEL; + } + if (mmcra & MMCRA_SIHV) + return PERF_EVENT_MISC_HYPERVISOR; + return (mmcra & MMCRA_SIPR) ? PERF_EVENT_MISC_USER : + PERF_EVENT_MISC_KERNEL; +} + +/* + * Overload regs->dsisr to store MMCRA so we only need to read it once + * on each interrupt. + */ +static inline void perf_read_regs(struct pt_regs *regs) +{ + regs->dsisr = mfspr(SPRN_MMCRA); +} + +/* + * If interrupts were soft-disabled when a PMU interrupt occurs, treat + * it as an NMI. + */ +static inline int perf_intr_is_nmi(struct pt_regs *regs) +{ + return !regs->softe; +} + +#endif /* CONFIG_PPC64 */ + static void perf_counter_interrupt(struct pt_regs *regs); void perf_counter_print_debug(void) @@ -78,12 +187,14 @@ static unsigned long read_pmc(int idx) case 6: val = mfspr(SPRN_PMC6); break; +#ifdef CONFIG_PPC64 case 7: val = mfspr(SPRN_PMC7); break; case 8: val = mfspr(SPRN_PMC8); break; +#endif /* CONFIG_PPC64 */ default: printk(KERN_ERR "oops trying to read PMC%d\n", idx); val = 0; @@ -115,12 +226,14 @@ static void write_pmc(int idx, unsigned long val) case 6: mtspr(SPRN_PMC6, val); break; +#ifdef CONFIG_PPC64 case 7: mtspr(SPRN_PMC7, val); break; case 8: mtspr(SPRN_PMC8, val); break; +#endif /* CONFIG_PPC64 */ default: printk(KERN_ERR "oops trying to write PMC%d\n", idx); } @@ -283,7 +396,7 @@ static int check_excludes(struct perf_counter **ctrs, unsigned int cflags[], static void power_pmu_read(struct perf_counter *counter) { - long val, delta, prev; + s64 val, delta, prev; if (!counter->hw.idx) return; @@ -477,7 +590,7 @@ void hw_perf_enable(void) mtspr(SPRN_MMCRA, cpuhw->mmcr[2] & ~MMCRA_SAMPLE_ENABLE); mtspr(SPRN_MMCR1, cpuhw->mmcr[1]); if (cpuhw->n_counters == 0) - get_lppaca()->pmcregs_in_use = 0; + perf_set_pmu_inuse(0); goto out_enable; } @@ -510,7 +623,7 @@ void hw_perf_enable(void) * bit set and set the hardware counters to their initial values. * Then unfreeze the counters. */ - get_lppaca()->pmcregs_in_use = 1; + perf_set_pmu_inuse(1); mtspr(SPRN_MMCRA, cpuhw->mmcr[2] & ~MMCRA_SAMPLE_ENABLE); mtspr(SPRN_MMCR1, cpuhw->mmcr[1]); mtspr(SPRN_MMCR0, (cpuhw->mmcr[0] & ~(MMCR0_PMC1CE | MMCR0_PMCjCE)) @@ -1007,11 +1120,10 @@ const struct pmu *hw_perf_counter_init(struct perf_counter *counter) * things if requested. Note that interrupts are hard-disabled * here so there is no possibility of being interrupted. */ -static void record_and_restart(struct perf_counter *counter, long val, +static void record_and_restart(struct perf_counter *counter, unsigned long val, struct pt_regs *regs, int nmi) { u64 period = counter->hw.sample_period; - unsigned long mmcra, sdsync; s64 prev, delta, left; int record = 0; @@ -1033,8 +1145,8 @@ static void record_and_restart(struct perf_counter *counter, long val, left = period; record = 1; } - if (left < 0x80000000L) - val = 0x80000000L - left; + if (left < 0x80000000LL) + val = 0x80000000LL - left; } /* @@ -1047,22 +1159,9 @@ static void record_and_restart(struct perf_counter *counter, long val, .period = counter->hw.last_period, }; - if (counter->attr.sample_type & PERF_SAMPLE_ADDR) { - /* - * The user wants a data address recorded. - * If we're not doing instruction sampling, - * give them the SDAR (sampled data address). - * If we are doing instruction sampling, then only - * give them the SDAR if it corresponds to the - * instruction pointed to by SIAR; this is indicated - * by the [POWER6_]MMCRA_SDSYNC bit in MMCRA. - */ - mmcra = regs->dsisr; - sdsync = (ppmu->flags & PPMU_ALT_SIPR) ? - POWER6_MMCRA_SDSYNC : MMCRA_SDSYNC; - if (!(mmcra & MMCRA_SAMPLE_ENABLE) || (mmcra & sdsync)) - data.addr = mfspr(SPRN_SDAR); - } + if (counter->attr.sample_type & PERF_SAMPLE_ADDR) + perf_get_data_addr(regs, &data.addr); + if (perf_counter_overflow(counter, nmi, &data)) { /* * Interrupts are coming too fast - throttle them @@ -1088,25 +1187,12 @@ static void record_and_restart(struct perf_counter *counter, long val, */ unsigned long perf_misc_flags(struct pt_regs *regs) { - unsigned long mmcra; + u32 flags = perf_get_misc_flags(regs); - if (TRAP(regs) != 0xf00) { - /* not a PMU interrupt */ - return user_mode(regs) ? PERF_EVENT_MISC_USER : - PERF_EVENT_MISC_KERNEL; - } - - mmcra = regs->dsisr; - if (ppmu->flags & PPMU_ALT_SIPR) { - if (mmcra & POWER6_MMCRA_SIHV) - return PERF_EVENT_MISC_HYPERVISOR; - return (mmcra & POWER6_MMCRA_SIPR) ? PERF_EVENT_MISC_USER : - PERF_EVENT_MISC_KERNEL; - } - if (mmcra & MMCRA_SIHV) - return PERF_EVENT_MISC_HYPERVISOR; - return (mmcra & MMCRA_SIPR) ? PERF_EVENT_MISC_USER : - PERF_EVENT_MISC_KERNEL; + if (flags) + return flags; + return user_mode(regs) ? PERF_EVENT_MISC_USER : + PERF_EVENT_MISC_KERNEL; } /* @@ -1115,20 +1201,12 @@ unsigned long perf_misc_flags(struct pt_regs *regs) */ unsigned long perf_instruction_pointer(struct pt_regs *regs) { - unsigned long mmcra; unsigned long ip; - unsigned long slot; if (TRAP(regs) != 0xf00) return regs->nip; /* not a PMU interrupt */ - ip = mfspr(SPRN_SIAR); - mmcra = regs->dsisr; - if ((mmcra & MMCRA_SAMPLE_ENABLE) && !(ppmu->flags & PPMU_ALT_SIPR)) { - slot = (mmcra & MMCRA_SLOT) >> MMCRA_SLOT_SHIFT; - if (slot > 1) - ip += 4 * (slot - 1); - } + ip = mfspr(SPRN_SIAR) + perf_ip_adjust(regs); return ip; } @@ -1140,7 +1218,7 @@ static void perf_counter_interrupt(struct pt_regs *regs) int i; struct cpu_hw_counters *cpuhw = &__get_cpu_var(cpu_hw_counters); struct perf_counter *counter; - long val; + unsigned long val; int found = 0; int nmi; @@ -1148,16 +1226,9 @@ static void perf_counter_interrupt(struct pt_regs *regs) freeze_limited_counters(cpuhw, mfspr(SPRN_PMC5), mfspr(SPRN_PMC6)); - /* - * Overload regs->dsisr to store MMCRA so we only need to read it once. - */ - regs->dsisr = mfspr(SPRN_MMCRA); + perf_read_regs(regs); - /* - * If interrupts were soft-disabled when this PMU interrupt - * occurred, treat it as an NMI. - */ - nmi = !regs->softe; + nmi = perf_intr_is_nmi(regs); if (nmi) nmi_enter(); else @@ -1223,11 +1294,13 @@ int register_power_pmu(struct power_pmu *pmu) pr_info("%s performance monitor hardware support registered\n", pmu->name); +#ifdef MSR_HV /* * Use FCHV to ignore kernel events if MSR.HV is set. */ if (mfmsr() & MSR_HV) freeze_counters_kernel = MMCR0_FCHV; +#endif /* CONFIG_PPC64 */ return 0; } -- cgit v1.2.3 From 7325927e5a20bfe0f006acf92801bf41c537d3d4 Mon Sep 17 00:00:00 2001 From: Paul Mackerras Date: Wed, 17 Jun 2009 21:53:51 +1000 Subject: perf_counter: powerpc: Add processor back-end for MPC7450 family This adds support for the performance monitor hardware on the MPC7450 family of processors (7450, 7451, 7455, 7447/7457, 7447A, 7448), used in the later Apple G4 powermacs/powerbooks and other machines. These machines have 6 hardware counters with a unique set of events which can be counted on each counter, with some events being available on multiple counters. Raw event codes for these processors are (PMC << 8) + PMCSEL. If PMC is non-zero then the event is that selected by the given PMCSEL value for that PMC (hardware counter). If PMC is zero then the event selected is one of the low-numbered ones that are common to several PMCs. In this case PMCSEL must be <= 22 and the event is what that PMCSEL value would select on PMC1 (but it may be placed any other PMC that has the same event for that PMCSEL value). For events that count cycles or occurrences that exceed a threshold, the threshold requested can be specified in the 0x3f000 bits of the raw event codes. If the event uses the threshold multiplier bit and that bit should be set, that is indicated with the 0x40000 bit of the raw event code. This fills in some of the generic cache events. Unfortunately there are quite a few blank spaces in the table, partly because these processors tend to count cache hits rather than cache accesses. Signed-off-by: Paul Mackerras Cc: Peter Zijlstra Cc: linuxppc-dev@ozlabs.org Cc: benh@kernel.crashing.org LKML-Reference: <19000.55631.802122.696927@cargo.ozlabs.ibm.com> Signed-off-by: Ingo Molnar --- arch/powerpc/kernel/Makefile | 2 + arch/powerpc/kernel/mpc7450-pmu.c | 417 +++++++++++++++++++++++++++++++++ arch/powerpc/platforms/Kconfig.cputype | 1 + 3 files changed, 420 insertions(+) create mode 100644 arch/powerpc/kernel/mpc7450-pmu.c (limited to 'arch') diff --git a/arch/powerpc/kernel/Makefile b/arch/powerpc/kernel/Makefile index c5f93f061927..a9f882963379 100644 --- a/arch/powerpc/kernel/Makefile +++ b/arch/powerpc/kernel/Makefile @@ -98,6 +98,7 @@ obj-$(CONFIG_FUNCTION_GRAPH_TRACER) += ftrace.o obj-$(CONFIG_PPC_PERF_CTRS) += perf_counter.o obj64-$(CONFIG_PPC_PERF_CTRS) += power4-pmu.o ppc970-pmu.o power5-pmu.o \ power5+-pmu.o power6-pmu.o power7-pmu.o +obj32-$(CONFIG_PPC_PERF_CTRS) += mpc7450-pmu.o obj-$(CONFIG_8XX_MINIMAL_FPEMU) += softemu8xx.o @@ -106,6 +107,7 @@ obj-y += iomap.o endif obj-$(CONFIG_PPC64) += $(obj64-y) +obj-$(CONFIG_PPC32) += $(obj32-y) ifneq ($(CONFIG_XMON)$(CONFIG_KEXEC),) obj-y += ppc_save_regs.o diff --git a/arch/powerpc/kernel/mpc7450-pmu.c b/arch/powerpc/kernel/mpc7450-pmu.c new file mode 100644 index 000000000000..75ff47fed7bf --- /dev/null +++ b/arch/powerpc/kernel/mpc7450-pmu.c @@ -0,0 +1,417 @@ +/* + * Performance counter support for MPC7450-family processors. + * + * Copyright 2008-2009 Paul Mackerras, IBM Corporation. + * + * This program is free software; you can redistribute it and/or + * modify it under the terms of the GNU General Public License + * as published by the Free Software Foundation; either version + * 2 of the License, or (at your option) any later version. + */ +#include +#include +#include +#include +#include + +#define N_COUNTER 6 /* Number of hardware counters */ +#define MAX_ALT 3 /* Maximum number of event alternative codes */ + +/* + * Bits in event code for MPC7450 family + */ +#define PM_THRMULT_MSKS 0x40000 +#define PM_THRESH_SH 12 +#define PM_THRESH_MSK 0x3f +#define PM_PMC_SH 8 +#define PM_PMC_MSK 7 +#define PM_PMCSEL_MSK 0x7f + +/* + * Classify events according to how specific their PMC requirements are. + * Result is: + * 0: can go on any PMC + * 1: can go on PMCs 1-4 + * 2: can go on PMCs 1,2,4 + * 3: can go on PMCs 1 or 2 + * 4: can only go on one PMC + * -1: event code is invalid + */ +#define N_CLASSES 5 + +static int mpc7450_classify_event(u32 event) +{ + int pmc; + + pmc = (event >> PM_PMC_SH) & PM_PMC_MSK; + if (pmc) { + if (pmc > N_COUNTER) + return -1; + return 4; + } + event &= PM_PMCSEL_MSK; + if (event <= 1) + return 0; + if (event <= 7) + return 1; + if (event <= 13) + return 2; + if (event <= 22) + return 3; + return -1; +} + +/* + * Events using threshold and possible threshold scale: + * code scale? name + * 11e N PM_INSTQ_EXCEED_CYC + * 11f N PM_ALTV_IQ_EXCEED_CYC + * 128 Y PM_DTLB_SEARCH_EXCEED_CYC + * 12b Y PM_LD_MISS_EXCEED_L1_CYC + * 220 N PM_CQ_EXCEED_CYC + * 30c N PM_GPR_RB_EXCEED_CYC + * 30d ? PM_FPR_IQ_EXCEED_CYC ? + * 311 Y PM_ITLB_SEARCH_EXCEED + * 410 N PM_GPR_IQ_EXCEED_CYC + */ + +/* + * Return use of threshold and threshold scale bits: + * 0 = uses neither, 1 = uses threshold, 2 = uses both + */ +static int mpc7450_threshold_use(u32 event) +{ + int pmc, sel; + + pmc = (event >> PM_PMC_SH) & PM_PMC_MSK; + sel = event & PM_PMCSEL_MSK; + switch (pmc) { + case 1: + if (sel == 0x1e || sel == 0x1f) + return 1; + if (sel == 0x28 || sel == 0x2b) + return 2; + break; + case 2: + if (sel == 0x20) + return 1; + break; + case 3: + if (sel == 0xc || sel == 0xd) + return 1; + if (sel == 0x11) + return 2; + break; + case 4: + if (sel == 0x10) + return 1; + break; + } + return 0; +} + +/* + * Layout of constraint bits: + * 33222222222211111111110000000000 + * 10987654321098765432109876543210 + * |< >< > < > < ><><><><><><> + * TS TV G4 G3 G2P6P5P4P3P2P1 + * + * P1 - P6 + * 0 - 11: Count of events needing PMC1 .. PMC6 + * + * G2 + * 12 - 14: Count of events needing PMC1 or PMC2 + * + * G3 + * 16 - 18: Count of events needing PMC1, PMC2 or PMC4 + * + * G4 + * 20 - 23: Count of events needing PMC1, PMC2, PMC3 or PMC4 + * + * TV + * 24 - 29: Threshold value requested + * + * TS + * 30: Threshold scale value requested + */ + +static u32 pmcbits[N_COUNTER][2] = { + { 0x00844002, 0x00111001 }, /* PMC1 mask, value: P1,G2,G3,G4 */ + { 0x00844008, 0x00111004 }, /* PMC2: P2,G2,G3,G4 */ + { 0x00800020, 0x00100010 }, /* PMC3: P3,G4 */ + { 0x00840080, 0x00110040 }, /* PMC4: P4,G3,G4 */ + { 0x00000200, 0x00000100 }, /* PMC5: P5 */ + { 0x00000800, 0x00000400 } /* PMC6: P6 */ +}; + +static u32 classbits[N_CLASSES - 1][2] = { + { 0x00000000, 0x00000000 }, /* class 0: no constraint */ + { 0x00800000, 0x00100000 }, /* class 1: G4 */ + { 0x00040000, 0x00010000 }, /* class 2: G3 */ + { 0x00004000, 0x00001000 }, /* class 3: G2 */ +}; + +static int mpc7450_get_constraint(u64 event, unsigned long *maskp, + unsigned long *valp) +{ + int pmc, class; + u32 mask, value; + int thresh, tuse; + + class = mpc7450_classify_event(event); + if (class < 0) + return -1; + if (class == 4) { + pmc = ((unsigned int)event >> PM_PMC_SH) & PM_PMC_MSK; + mask = pmcbits[pmc - 1][0]; + value = pmcbits[pmc - 1][1]; + } else { + mask = classbits[class][0]; + value = classbits[class][1]; + } + + tuse = mpc7450_threshold_use(event); + if (tuse) { + thresh = ((unsigned int)event >> PM_THRESH_SH) & PM_THRESH_MSK; + mask |= 0x3f << 24; + value |= thresh << 24; + if (tuse == 2) { + mask |= 0x40000000; + if ((unsigned int)event & PM_THRMULT_MSKS) + value |= 0x40000000; + } + } + + *maskp = mask; + *valp = value; + return 0; +} + +static const unsigned int event_alternatives[][MAX_ALT] = { + { 0x217, 0x317 }, /* PM_L1_DCACHE_MISS */ + { 0x418, 0x50f, 0x60f }, /* PM_SNOOP_RETRY */ + { 0x502, 0x602 }, /* PM_L2_HIT */ + { 0x503, 0x603 }, /* PM_L3_HIT */ + { 0x504, 0x604 }, /* PM_L2_ICACHE_MISS */ + { 0x505, 0x605 }, /* PM_L3_ICACHE_MISS */ + { 0x506, 0x606 }, /* PM_L2_DCACHE_MISS */ + { 0x507, 0x607 }, /* PM_L3_DCACHE_MISS */ + { 0x50a, 0x623 }, /* PM_LD_HIT_L3 */ + { 0x50b, 0x624 }, /* PM_ST_HIT_L3 */ + { 0x50d, 0x60d }, /* PM_L2_TOUCH_HIT */ + { 0x50e, 0x60e }, /* PM_L3_TOUCH_HIT */ + { 0x512, 0x612 }, /* PM_INT_LOCAL */ + { 0x513, 0x61d }, /* PM_L2_MISS */ + { 0x514, 0x61e }, /* PM_L3_MISS */ +}; + +/* + * Scan the alternatives table for a match and return the + * index into the alternatives table if found, else -1. + */ +static int find_alternative(u32 event) +{ + int i, j; + + for (i = 0; i < ARRAY_SIZE(event_alternatives); ++i) { + if (event < event_alternatives[i][0]) + break; + for (j = 0; j < MAX_ALT && event_alternatives[i][j]; ++j) + if (event == event_alternatives[i][j]) + return i; + } + return -1; +} + +static int mpc7450_get_alternatives(u64 event, unsigned int flags, u64 alt[]) +{ + int i, j, nalt = 1; + u32 ae; + + alt[0] = event; + nalt = 1; + i = find_alternative((u32)event); + if (i >= 0) { + for (j = 0; j < MAX_ALT; ++j) { + ae = event_alternatives[i][j]; + if (ae && ae != (u32)event) + alt[nalt++] = ae; + } + } + return nalt; +} + +/* + * Bitmaps of which PMCs each class can use for classes 0 - 3. + * Bit i is set if PMC i+1 is usable. + */ +static const u8 classmap[N_CLASSES] = { + 0x3f, 0x0f, 0x0b, 0x03, 0 +}; + +/* Bit position and width of each PMCSEL field */ +static const int pmcsel_shift[N_COUNTER] = { + 6, 0, 27, 22, 17, 11 +}; +static const u32 pmcsel_mask[N_COUNTER] = { + 0x7f, 0x3f, 0x1f, 0x1f, 0x1f, 0x3f +}; + +/* + * Compute MMCR0/1/2 values for a set of events. + */ +static int mpc7450_compute_mmcr(u64 event[], int n_ev, + unsigned int hwc[], unsigned long mmcr[]) +{ + u8 event_index[N_CLASSES][N_COUNTER]; + int n_classevent[N_CLASSES]; + int i, j, class, tuse; + u32 pmc_inuse = 0, pmc_avail; + u32 mmcr0 = 0, mmcr1 = 0, mmcr2 = 0; + u32 ev, pmc, thresh; + + if (n_ev > N_COUNTER) + return -1; + + /* First pass: count usage in each class */ + for (i = 0; i < N_CLASSES; ++i) + n_classevent[i] = 0; + for (i = 0; i < n_ev; ++i) { + class = mpc7450_classify_event(event[i]); + if (class < 0) + return -1; + j = n_classevent[class]++; + event_index[class][j] = i; + } + + /* Second pass: allocate PMCs from most specific event to least */ + for (class = N_CLASSES - 1; class >= 0; --class) { + for (i = 0; i < n_classevent[class]; ++i) { + ev = event[event_index[class][i]]; + if (class == 4) { + pmc = (ev >> PM_PMC_SH) & PM_PMC_MSK; + if (pmc_inuse & (1 << (pmc - 1))) + return -1; + } else { + /* Find a suitable PMC */ + pmc_avail = classmap[class] & ~pmc_inuse; + if (!pmc_avail) + return -1; + pmc = ffs(pmc_avail); + } + pmc_inuse |= 1 << (pmc - 1); + + tuse = mpc7450_threshold_use(ev); + if (tuse) { + thresh = (ev >> PM_THRESH_SH) & PM_THRESH_MSK; + mmcr0 |= thresh << 16; + if (tuse == 2 && (ev & PM_THRMULT_MSKS)) + mmcr2 = 0x80000000; + } + ev &= pmcsel_mask[pmc - 1]; + ev <<= pmcsel_shift[pmc - 1]; + if (pmc <= 2) + mmcr0 |= ev; + else + mmcr1 |= ev; + hwc[event_index[class][i]] = pmc - 1; + } + } + + if (pmc_inuse & 1) + mmcr0 |= MMCR0_PMC1CE; + if (pmc_inuse & 0x3e) + mmcr0 |= MMCR0_PMCnCE; + + /* Return MMCRx values */ + mmcr[0] = mmcr0; + mmcr[1] = mmcr1; + mmcr[2] = mmcr2; + return 0; +} + +/* + * Disable counting by a PMC. + * Note that the pmc argument is 0-based here, not 1-based. + */ +static void mpc7450_disable_pmc(unsigned int pmc, unsigned long mmcr[]) +{ + if (pmc <= 1) + mmcr[0] &= ~(pmcsel_mask[pmc] << pmcsel_shift[pmc]); + else + mmcr[1] &= ~(pmcsel_mask[pmc] << pmcsel_shift[pmc]); +} + +static int mpc7450_generic_events[] = { + [PERF_COUNT_HW_CPU_CYCLES] = 1, + [PERF_COUNT_HW_INSTRUCTIONS] = 2, + [PERF_COUNT_HW_CACHE_MISSES] = 0x217, /* PM_L1_DCACHE_MISS */ + [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = 0x122, /* PM_BR_CMPL */ + [PERF_COUNT_HW_BRANCH_MISSES] = 0x41c, /* PM_BR_MPRED */ +}; + +#define C(x) PERF_COUNT_HW_CACHE_##x + +/* + * Table of generalized cache-related events. + * 0 means not supported, -1 means nonsensical, other values + * are event codes. + */ +static int mpc7450_cache_events[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = { + [C(L1D)] = { /* RESULT_ACCESS RESULT_MISS */ + [C(OP_READ)] = { 0, 0x225 }, + [C(OP_WRITE)] = { 0, 0x227 }, + [C(OP_PREFETCH)] = { 0, 0 }, + }, + [C(L1I)] = { /* RESULT_ACCESS RESULT_MISS */ + [C(OP_READ)] = { 0x129, 0x115 }, + [C(OP_WRITE)] = { -1, -1 }, + [C(OP_PREFETCH)] = { 0x634, 0 }, + }, + [C(LL)] = { /* RESULT_ACCESS RESULT_MISS */ + [C(OP_READ)] = { 0, 0 }, + [C(OP_WRITE)] = { 0, 0 }, + [C(OP_PREFETCH)] = { 0, 0 }, + }, + [C(DTLB)] = { /* RESULT_ACCESS RESULT_MISS */ + [C(OP_READ)] = { 0, 0x312 }, + [C(OP_WRITE)] = { -1, -1 }, + [C(OP_PREFETCH)] = { -1, -1 }, + }, + [C(ITLB)] = { /* RESULT_ACCESS RESULT_MISS */ + [C(OP_READ)] = { 0, 0x223 }, + [C(OP_WRITE)] = { -1, -1 }, + [C(OP_PREFETCH)] = { -1, -1 }, + }, + [C(BPU)] = { /* RESULT_ACCESS RESULT_MISS */ + [C(OP_READ)] = { 0x122, 0x41c }, + [C(OP_WRITE)] = { -1, -1 }, + [C(OP_PREFETCH)] = { -1, -1 }, + }, +}; + +struct power_pmu mpc7450_pmu = { + .name = "MPC7450 family", + .n_counter = N_COUNTER, + .max_alternatives = MAX_ALT, + .add_fields = 0x00111555ul, + .test_adder = 0x00301000ul, + .compute_mmcr = mpc7450_compute_mmcr, + .get_constraint = mpc7450_get_constraint, + .get_alternatives = mpc7450_get_alternatives, + .disable_pmc = mpc7450_disable_pmc, + .n_generic = ARRAY_SIZE(mpc7450_generic_events), + .generic_events = mpc7450_generic_events, + .cache_events = &mpc7450_cache_events, +}; + +static int init_mpc7450_pmu(void) +{ + if (strcmp(cur_cpu_spec->oprofile_cpu_type, "ppc/7450")) + return -ENODEV; + + return register_power_pmu(&mpc7450_pmu); +} + +arch_initcall(init_mpc7450_pmu); diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype index dd9f3ec5ee30..8485c8ca7a06 100644 --- a/arch/powerpc/platforms/Kconfig.cputype +++ b/arch/powerpc/platforms/Kconfig.cputype @@ -75,6 +75,7 @@ config POWER4_ONLY config 6xx def_bool y depends on PPC32 && PPC_BOOK3S + select PPC_HAVE_PMU_SUPPORT config POWER3 bool -- cgit v1.2.3 From f9188e023c248d73f5b4a589b480e065c1864068 Mon Sep 17 00:00:00 2001 From: Peter Zijlstra Date: Thu, 18 Jun 2009 22:20:52 +0200 Subject: perf_counter: Make callchain samples extensible Before exposing upstream tools to a callchain-samples ABI, tidy it up to make it more extensible in the future: Use markers in the IP chain to denote context, use (u64)-1..-4095 range for these context markers because we use them for ERR_PTR(), so these addresses are unlikely to be mapped. Signed-off-by: Peter Zijlstra Cc: Mike Galbraith Cc: Paul Mackerras Cc: Arnaldo Carvalho de Melo LKML-Reference: Signed-off-by: Ingo Molnar --- arch/x86/kernel/cpu/perf_counter.c | 29 ++++++----------------------- include/linux/perf_counter.h | 28 +++++++++++++++++----------- 2 files changed, 23 insertions(+), 34 deletions(-) (limited to 'arch') diff --git a/arch/x86/kernel/cpu/perf_counter.c b/arch/x86/kernel/cpu/perf_counter.c index ce1ae3f1f86c..76dfef23f789 100644 --- a/arch/x86/kernel/cpu/perf_counter.c +++ b/arch/x86/kernel/cpu/perf_counter.c @@ -1555,9 +1555,9 @@ const struct pmu *hw_perf_counter_init(struct perf_counter *counter) */ static inline -void callchain_store(struct perf_callchain_entry *entry, unsigned long ip) +void callchain_store(struct perf_callchain_entry *entry, u64 ip) { - if (entry->nr < MAX_STACK_DEPTH) + if (entry->nr < PERF_MAX_STACK_DEPTH) entry->ip[entry->nr++] = ip; } @@ -1602,22 +1602,10 @@ static const struct stacktrace_ops backtrace_ops = { static void perf_callchain_kernel(struct pt_regs *regs, struct perf_callchain_entry *entry) { - unsigned long bp; - char *stack; - int nr = entry->nr; - + callchain_store(entry, PERF_CONTEXT_KERNEL); callchain_store(entry, regs->ip); - stack = ((char *)regs + sizeof(struct pt_regs)); -#ifdef CONFIG_FRAME_POINTER - get_bp(bp); -#else - bp = 0; -#endif - - dump_trace(NULL, regs, (void *)&stack, bp, &backtrace_ops, entry); - - entry->kernel = entry->nr - nr; + dump_trace(NULL, regs, NULL, 0, &backtrace_ops, entry); } /* @@ -1669,16 +1657,16 @@ perf_callchain_user(struct pt_regs *regs, struct perf_callchain_entry *entry) { struct stack_frame frame; const void __user *fp; - int nr = entry->nr; if (!user_mode(regs)) regs = task_pt_regs(current); fp = (void __user *)regs->bp; + callchain_store(entry, PERF_CONTEXT_USER); callchain_store(entry, regs->ip); - while (entry->nr < MAX_STACK_DEPTH) { + while (entry->nr < PERF_MAX_STACK_DEPTH) { frame.next_frame = NULL; frame.return_address = 0; @@ -1691,8 +1679,6 @@ perf_callchain_user(struct pt_regs *regs, struct perf_callchain_entry *entry) callchain_store(entry, frame.return_address); fp = frame.next_frame; } - - entry->user = entry->nr - nr; } static void @@ -1728,9 +1714,6 @@ struct perf_callchain_entry *perf_callchain(struct pt_regs *regs) entry = &__get_cpu_var(irq_entry); entry->nr = 0; - entry->hv = 0; - entry->kernel = 0; - entry->user = 0; perf_do_callchain(regs, entry); diff --git a/include/linux/perf_counter.h b/include/linux/perf_counter.h index 0765e8e69843..e7e7e0242767 100644 --- a/include/linux/perf_counter.h +++ b/include/linux/perf_counter.h @@ -343,23 +343,22 @@ enum perf_event_type { * { u64 nr; * { u64 id, val; } cnt[nr]; } && PERF_SAMPLE_GROUP * - * { u16 nr, - * hv, - * kernel, - * user; + * { u64 nr, * u64 ips[nr]; } && PERF_SAMPLE_CALLCHAIN * }; */ }; -#define MAX_STACK_DEPTH 255 +enum perf_callchain_context { + PERF_CONTEXT_HV = (__u64)-32, + PERF_CONTEXT_KERNEL = (__u64)-128, + PERF_CONTEXT_USER = (__u64)-512, -struct perf_callchain_entry { - __u16 nr; - __u16 hv; - __u16 kernel; - __u16 user; - __u64 ip[MAX_STACK_DEPTH]; + PERF_CONTEXT_GUEST = (__u64)-2048, + PERF_CONTEXT_GUEST_KERNEL = (__u64)-2176, + PERF_CONTEXT_GUEST_USER = (__u64)-2560, + + PERF_CONTEXT_MAX = (__u64)-4095, }; #ifdef __KERNEL__ @@ -381,6 +380,13 @@ struct perf_callchain_entry { #include #include +#define PERF_MAX_STACK_DEPTH 255 + +struct perf_callchain_entry { + __u64 nr; + __u64 ip[PERF_MAX_STACK_DEPTH]; +}; + struct task_struct; /** -- cgit v1.2.3 From 0c87197142427063e096f11603543ca874045952 Mon Sep 17 00:00:00 2001 From: Ingo Molnar Date: Mon, 15 Jun 2009 11:35:01 +0200 Subject: perf_counter, x86: Improve interactions with fast-gup Improve a few details in perfcounter call-chain recording that makes use of fast-GUP: - Use ACCESS_ONCE() to observe the pte value. ptes are fundamentally racy and can be changed on another CPU, so we have to be careful about how we access them. The PAE branch is already careful with read-barriers - but the non-PAE and 64-bit side needs an ACCESS_ONCE() to make sure the pte value is observed only once. - make the checks a bit stricter so that we can feed it any kind of cra^H^H^H user-space input ;-) Acked-by: Peter Zijlstra Signed-off-by: Ingo Molnar --- arch/x86/include/asm/uaccess.h | 7 ++++++- arch/x86/mm/gup.c | 2 +- 2 files changed, 7 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/x86/include/asm/uaccess.h b/arch/x86/include/asm/uaccess.h index b685ece89d5c..512ee87062c2 100644 --- a/arch/x86/include/asm/uaccess.h +++ b/arch/x86/include/asm/uaccess.h @@ -25,7 +25,12 @@ #define MAKE_MM_SEG(s) ((mm_segment_t) { (s) }) #define KERNEL_DS MAKE_MM_SEG(-1UL) -#define USER_DS MAKE_MM_SEG(PAGE_OFFSET) + +#ifdef CONFIG_X86_32 +# define USER_DS MAKE_MM_SEG(PAGE_OFFSET) +#else +# define USER_DS MAKE_MM_SEG(__VIRTUAL_MASK) +#endif #define get_ds() (KERNEL_DS) #define get_fs() (current_thread_info()->addr_limit) diff --git a/arch/x86/mm/gup.c b/arch/x86/mm/gup.c index 697d5727c119..2d1d784ad3f7 100644 --- a/arch/x86/mm/gup.c +++ b/arch/x86/mm/gup.c @@ -14,7 +14,7 @@ static inline pte_t gup_get_pte(pte_t *ptep) { #ifndef CONFIG_X86_PAE - return *ptep; + return ACCESS_ONCE(*ptep); #else /* * With get_user_pages_fast, we walk down the pagetables without taking -- cgit v1.2.3