* are redone after IRQs are re-enabled (and gcc doesn't reuse stale register)
*
* Noted at the time of Abilis Timer List corruption
- * Orig Bug + Rejected solution : https://lkml.org/lkml/2013/3/29/67
- * Reasoning : https://lkml.org/lkml/2013/4/8/15
+ *
+ * Orig Bug + Rejected solution:
+ * https://lore.kernel.org/lkml/1364553218-31255-1-git-send-email-vgupta@synopsys.com
+ *
+ * Reasoning:
+ * https://lore.kernel.org/lkml/CA+55aFyFWjpSVQM6M266tKrG_ZXJzZ-nYejpmXYQXbrr42mGPQ@mail.gmail.com
*
******************************************************************/
#define PERF_COUNT_ARC_HW_MAX (PERF_COUNT_HW_MAX + 8)
-/*
- * Some ARC pct quirks:
- *
- * PERF_COUNT_HW_STALLED_CYCLES_BACKEND
- * PERF_COUNT_HW_STALLED_CYCLES_FRONTEND
- * The ARC 700 can either measure stalls per pipeline stage, or all stalls
- * combined; for now we assign all stalls to STALLED_CYCLES_BACKEND
- * and all pipeline flushes (e.g. caused by mispredicts, etc.) to
- * STALLED_CYCLES_FRONTEND.
- *
- * We could start multiple performance counters and combine everything
- * afterwards, but that makes it complicated.
- *
- * Note that I$ cache misses aren't counted by either of the two!
- */
-
-/*
- * ARC PCT has hardware conditions with fixed "names" but variable "indexes"
- * (based on a specific RTL build)
- * Below is the static map between perf generic/arc specific event_id and
- * h/w condition names.
- * At the time of probe, we loop thru each index and find it's name to
- * complete the mapping of perf event_id to h/w index as latter is needed
- * to program the counter really
- */
-static const char * const arc_pmu_ev_hw_map[] = {
- /* count cycles */
- [PERF_COUNT_HW_CPU_CYCLES] = "crun",
- [PERF_COUNT_HW_REF_CPU_CYCLES] = "crun",
- [PERF_COUNT_HW_BUS_CYCLES] = "crun",
-
- [PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = "bflush",
- [PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = "bstall",
-
- /* counts condition */
- [PERF_COUNT_HW_INSTRUCTIONS] = "iall",
- /* All jump instructions that are taken */
- [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = "ijmptak",
-#ifdef CONFIG_ISA_ARCV2
- [PERF_COUNT_HW_BRANCH_MISSES] = "bpmp",
-#else
- [PERF_COUNT_ARC_BPOK] = "bpok", /* NP-NT, PT-T, PNT-NT */
- [PERF_COUNT_HW_BRANCH_MISSES] = "bpfail", /* NP-T, PT-NT, PNT-T */
-#endif
- [PERF_COUNT_ARC_LDC] = "imemrdc", /* Instr: mem read cached */
- [PERF_COUNT_ARC_STC] = "imemwrc", /* Instr: mem write cached */
-
- [PERF_COUNT_ARC_DCLM] = "dclm", /* D-cache Load Miss */
- [PERF_COUNT_ARC_DCSM] = "dcsm", /* D-cache Store Miss */
- [PERF_COUNT_ARC_ICM] = "icm", /* I-cache Miss */
- [PERF_COUNT_ARC_EDTLB] = "edtlb", /* D-TLB Miss */
- [PERF_COUNT_ARC_EITLB] = "eitlb", /* I-TLB Miss */
-
- [PERF_COUNT_HW_CACHE_REFERENCES] = "imemrdc", /* Instr: mem read cached */
- [PERF_COUNT_HW_CACHE_MISSES] = "dclm", /* D-cache Load Miss */
-};
-
-#define C(_x) PERF_COUNT_HW_CACHE_##_x
-#define CACHE_OP_UNSUPPORTED 0xffff
-
-static const unsigned int arc_pmu_cache_map[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = {
- [C(L1D)] = {
- [C(OP_READ)] = {
- [C(RESULT_ACCESS)] = PERF_COUNT_ARC_LDC,
- [C(RESULT_MISS)] = PERF_COUNT_ARC_DCLM,
- },
- [C(OP_WRITE)] = {
- [C(RESULT_ACCESS)] = PERF_COUNT_ARC_STC,
- [C(RESULT_MISS)] = PERF_COUNT_ARC_DCSM,
- },
- [C(OP_PREFETCH)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- },
- [C(L1I)] = {
- [C(OP_READ)] = {
- [C(RESULT_ACCESS)] = PERF_COUNT_HW_INSTRUCTIONS,
- [C(RESULT_MISS)] = PERF_COUNT_ARC_ICM,
- },
- [C(OP_WRITE)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- [C(OP_PREFETCH)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- },
- [C(LL)] = {
- [C(OP_READ)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- [C(OP_WRITE)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- [C(OP_PREFETCH)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- },
- [C(DTLB)] = {
- [C(OP_READ)] = {
- [C(RESULT_ACCESS)] = PERF_COUNT_ARC_LDC,
- [C(RESULT_MISS)] = PERF_COUNT_ARC_EDTLB,
- },
- /* DTLB LD/ST Miss not segregated by h/w*/
- [C(OP_WRITE)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- [C(OP_PREFETCH)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- },
- [C(ITLB)] = {
- [C(OP_READ)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = PERF_COUNT_ARC_EITLB,
- },
- [C(OP_WRITE)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- [C(OP_PREFETCH)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- },
- [C(BPU)] = {
- [C(OP_READ)] = {
- [C(RESULT_ACCESS)] = PERF_COUNT_HW_BRANCH_INSTRUCTIONS,
- [C(RESULT_MISS)] = PERF_COUNT_HW_BRANCH_MISSES,
- },
- [C(OP_WRITE)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- [C(OP_PREFETCH)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- },
- [C(NODE)] = {
- [C(OP_READ)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- [C(OP_WRITE)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- [C(OP_PREFETCH)] = {
- [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
- [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
- },
- },
-};
-
#endif /* __ASM_PERF_EVENT_H */
/*
* _TIF_ALLWORK_MASK includes SYSCALL_TRACE, but we don't need it.
- * SYSCALL_TRACE is anyway seperately/unconditionally tested right after a
- * syscall, so all that reamins to be tested is _TIF_WORK_MASK
+ * SYSCALL_TRACE is anyway separately/unconditionally tested right after a
+ * syscall, so all that remains to be tested is _TIF_WORK_MASK
*/
#endif /* _ASM_THREAD_INFO_H */
/* HW holds 8 symbols + one for null terminator */
#define ARCPMU_EVENT_NAME_LEN 9
+/*
+ * Some ARC pct quirks:
+ *
+ * PERF_COUNT_HW_STALLED_CYCLES_BACKEND
+ * PERF_COUNT_HW_STALLED_CYCLES_FRONTEND
+ * The ARC 700 can either measure stalls per pipeline stage, or all stalls
+ * combined; for now we assign all stalls to STALLED_CYCLES_BACKEND
+ * and all pipeline flushes (e.g. caused by mispredicts, etc.) to
+ * STALLED_CYCLES_FRONTEND.
+ *
+ * We could start multiple performance counters and combine everything
+ * afterwards, but that makes it complicated.
+ *
+ * Note that I$ cache misses aren't counted by either of the two!
+ */
+
+/*
+ * ARC PCT has hardware conditions with fixed "names" but variable "indexes"
+ * (based on a specific RTL build)
+ * Below is the static map between perf generic/arc specific event_id and
+ * h/w condition names.
+ * At the time of probe, we loop thru each index and find it's name to
+ * complete the mapping of perf event_id to h/w index as latter is needed
+ * to program the counter really
+ */
+static const char * const arc_pmu_ev_hw_map[] = {
+ /* count cycles */
+ [PERF_COUNT_HW_CPU_CYCLES] = "crun",
+ [PERF_COUNT_HW_REF_CPU_CYCLES] = "crun",
+ [PERF_COUNT_HW_BUS_CYCLES] = "crun",
+
+ [PERF_COUNT_HW_STALLED_CYCLES_FRONTEND] = "bflush",
+ [PERF_COUNT_HW_STALLED_CYCLES_BACKEND] = "bstall",
+
+ /* counts condition */
+ [PERF_COUNT_HW_INSTRUCTIONS] = "iall",
+ /* All jump instructions that are taken */
+ [PERF_COUNT_HW_BRANCH_INSTRUCTIONS] = "ijmptak",
+#ifdef CONFIG_ISA_ARCV2
+ [PERF_COUNT_HW_BRANCH_MISSES] = "bpmp",
+#else
+ [PERF_COUNT_ARC_BPOK] = "bpok", /* NP-NT, PT-T, PNT-NT */
+ [PERF_COUNT_HW_BRANCH_MISSES] = "bpfail", /* NP-T, PT-NT, PNT-T */
+#endif
+ [PERF_COUNT_ARC_LDC] = "imemrdc", /* Instr: mem read cached */
+ [PERF_COUNT_ARC_STC] = "imemwrc", /* Instr: mem write cached */
+
+ [PERF_COUNT_ARC_DCLM] = "dclm", /* D-cache Load Miss */
+ [PERF_COUNT_ARC_DCSM] = "dcsm", /* D-cache Store Miss */
+ [PERF_COUNT_ARC_ICM] = "icm", /* I-cache Miss */
+ [PERF_COUNT_ARC_EDTLB] = "edtlb", /* D-TLB Miss */
+ [PERF_COUNT_ARC_EITLB] = "eitlb", /* I-TLB Miss */
+
+ [PERF_COUNT_HW_CACHE_REFERENCES] = "imemrdc", /* Instr: mem read cached */
+ [PERF_COUNT_HW_CACHE_MISSES] = "dclm", /* D-cache Load Miss */
+};
+
+#define C(_x) PERF_COUNT_HW_CACHE_##_x
+#define CACHE_OP_UNSUPPORTED 0xffff
+
+static const unsigned int arc_pmu_cache_map[C(MAX)][C(OP_MAX)][C(RESULT_MAX)] = {
+ [C(L1D)] = {
+ [C(OP_READ)] = {
+ [C(RESULT_ACCESS)] = PERF_COUNT_ARC_LDC,
+ [C(RESULT_MISS)] = PERF_COUNT_ARC_DCLM,
+ },
+ [C(OP_WRITE)] = {
+ [C(RESULT_ACCESS)] = PERF_COUNT_ARC_STC,
+ [C(RESULT_MISS)] = PERF_COUNT_ARC_DCSM,
+ },
+ [C(OP_PREFETCH)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ },
+ [C(L1I)] = {
+ [C(OP_READ)] = {
+ [C(RESULT_ACCESS)] = PERF_COUNT_HW_INSTRUCTIONS,
+ [C(RESULT_MISS)] = PERF_COUNT_ARC_ICM,
+ },
+ [C(OP_WRITE)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ [C(OP_PREFETCH)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ },
+ [C(LL)] = {
+ [C(OP_READ)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ [C(OP_WRITE)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ [C(OP_PREFETCH)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ },
+ [C(DTLB)] = {
+ [C(OP_READ)] = {
+ [C(RESULT_ACCESS)] = PERF_COUNT_ARC_LDC,
+ [C(RESULT_MISS)] = PERF_COUNT_ARC_EDTLB,
+ },
+ /* DTLB LD/ST Miss not segregated by h/w*/
+ [C(OP_WRITE)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ [C(OP_PREFETCH)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ },
+ [C(ITLB)] = {
+ [C(OP_READ)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = PERF_COUNT_ARC_EITLB,
+ },
+ [C(OP_WRITE)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ [C(OP_PREFETCH)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ },
+ [C(BPU)] = {
+ [C(OP_READ)] = {
+ [C(RESULT_ACCESS)] = PERF_COUNT_HW_BRANCH_INSTRUCTIONS,
+ [C(RESULT_MISS)] = PERF_COUNT_HW_BRANCH_MISSES,
+ },
+ [C(OP_WRITE)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ [C(OP_PREFETCH)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ },
+ [C(NODE)] = {
+ [C(OP_READ)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ [C(OP_WRITE)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ [C(OP_PREFETCH)] = {
+ [C(RESULT_ACCESS)] = CACHE_OP_UNSUPPORTED,
+ [C(RESULT_MISS)] = CACHE_OP_UNSUPPORTED,
+ },
+ },
+};
+
enum arc_pmu_attr_groups {
ARCPMU_ATTR_GR_EVENTS,
ARCPMU_ATTR_GR_FORMATS,
}
if (!(event->hw.state & PERF_HES_STOPPED)) {
- /* stop ARC pmu here */
+ /* stop hw counter here */
write_aux_reg(ARC_REG_PCT_INDEX, idx);
/* condition code #0 is always "never" */
{
struct arc_pmu_cpu *pmu_cpu = this_cpu_ptr(&arc_pmu_cpu);
struct hw_perf_event *hwc = &event->hw;
- int idx = hwc->idx;
+ int idx;
idx = ffz(pmu_cpu->used_mask[0]);
if (idx == arc_pmu->n_counters)
{
struct eh_frame_hdr_table_entry *e1 = p1;
struct eh_frame_hdr_table_entry *e2 = p2;
- unsigned long v;
-
- v = e1->start;
- e1->start = e2->start;
- e2->start = v;
- v = e1->fde;
- e1->fde = e2->fde;
- e2->fde = v;
+
+ swap(e1->start, e2->start);
+ swap(e1->fde, e2->fde);
}
static void init_unwind_hdr(struct unwind_table *table,
/*
* Cache operations depending on function and direction argument, inspired by
- * https://lkml.org/lkml/2018/5/18/979
+ * https://lore.kernel.org/lkml/20180518175004.GF17671@n2100.armlinux.org.uk
* "dma_sync_*_for_cpu and direction=TO_DEVICE (was Re: [PATCH 02/20]
* dma-mapping: provide a generic dma-noncoherent implementation)"
*
* Current implementation of "irq-dw-apb-ictl" driver doesn't work well
* with stacked INTCs. In particular problem happens if its master INTC
* not yet instantiated. See discussion here -
- * https://lkml.org/lkml/2015/3/4/755
+ * https://lore.kernel.org/lkml/54F6FE2C.7020309@synopsys.com
*
* So setup the first gpio block as a passive pass thru and hide it from
* DT hardware topology - connect MB intc directly to cpu intc
* Current implementation of "irq-dw-apb-ictl" driver doesn't work well
* with stacked INTCs. In particular problem happens if its master INTC
* not yet instantiated. See discussion here -
- * https://lkml.org/lkml/2015/3/4/755
+ * https://lore.kernel.org/lkml/54F6FE2C.7020309@synopsys.com
*
* So setup the first gpio block as a passive pass thru and hide it from
* DT hardware topology - connect intc directly to cpu intc
projects / platform / kernel / linux-rpi.git / commitdiff