From: Alexander Shishkin <alexander.shishkin@linux.intel.com>
To: Peter Zijlstra <peterz@infradead.org>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>,
Ingo Molnar <mingo@redhat.com>,
linux-kernel@vger.kernel.org, jolsa@redhat.com,
adrian.hunter@intel.com, mathieu.poirier@linaro.org,
mark.rutland@arm.com,
Alexander Shishkin <alexander.shishkin@linux.intel.com>
Subject: [PATCH v2 4/4] perf/x86/intel/pt: Opportunistically use single range output mode
Date: Tue, 22 Oct 2019 12:58:12 +0300 [thread overview]
Message-ID: <20191022095812.67071-5-alexander.shishkin@linux.intel.com> (raw)
In-Reply-To: <20191022095812.67071-1-alexander.shishkin@linux.intel.com>
Most of PT implementations support Single Range Output mode, which is
an alternative to ToPA that can be used for a single contiguous buffer
and if we don't require an interrupt, that is, in AUX snapshot mode.
Now that perf core will use high order allocations for the AUX buffer,
in many cases the first condition will also be satisfied.
The two most obvious benefits of the Single Range Output mode over the
ToPA are:
* not having to allocate the ToPA table(s),
* not using the ToPA walk hardware.
Make use of this functionality where available and appropriate.
Signed-off-by: Alexander Shishkin <alexander.shishkin@linux.intel.com>
---
arch/x86/events/intel/pt.c | 116 ++++++++++++++++++++++++++++---------
arch/x86/events/intel/pt.h | 2 +
2 files changed, 90 insertions(+), 28 deletions(-)
diff --git a/arch/x86/events/intel/pt.c b/arch/x86/events/intel/pt.c
index 2f20d5a333c1..6edd7b785861 100644
--- a/arch/x86/events/intel/pt.c
+++ b/arch/x86/events/intel/pt.c
@@ -491,7 +491,9 @@ static void pt_config(struct perf_event *event)
}
reg = pt_config_filters(event);
- reg |= RTIT_CTL_TOPA | RTIT_CTL_TRACEEN;
+ reg |= RTIT_CTL_TRACEEN;
+ if (!buf->single)
+ reg |= RTIT_CTL_TOPA;
/*
* Previously, we had BRANCH_EN on by default, but now that PT has
@@ -543,18 +545,6 @@ static void pt_config_stop(struct perf_event *event)
wmb();
}
-static void pt_config_buffer(void *buf, unsigned int topa_idx,
- unsigned int output_off)
-{
- u64 reg;
-
- wrmsrl(MSR_IA32_RTIT_OUTPUT_BASE, virt_to_phys(buf));
-
- reg = 0x7f | ((u64)topa_idx << 7) | ((u64)output_off << 32);
-
- wrmsrl(MSR_IA32_RTIT_OUTPUT_MASK, reg);
-}
-
/**
* struct topa - ToPA metadata
* @list: linkage to struct pt_buffer's list of tables
@@ -612,6 +602,26 @@ static inline phys_addr_t topa_pfn(struct topa *topa)
#define TOPA_ENTRY_SIZE(t, i) (sizes(TOPA_ENTRY((t), (i))->size))
#define TOPA_ENTRY_PAGES(t, i) (1 << TOPA_ENTRY((t), (i))->size)
+static void pt_config_buffer(struct pt_buffer *buf)
+{
+ u64 reg, mask;
+ void *base;
+
+ if (buf->single) {
+ base = buf->data_pages[0];
+ mask = (buf->nr_pages * PAGE_SIZE - 1) >> 7;
+ } else {
+ base = topa_to_page(buf->cur)->table;
+ mask = (u64)buf->cur_idx;
+ }
+
+ wrmsrl(MSR_IA32_RTIT_OUTPUT_BASE, virt_to_phys(base));
+
+ reg = 0x7f | (mask << 7) | ((u64)buf->output_off << 32);
+
+ wrmsrl(MSR_IA32_RTIT_OUTPUT_MASK, reg);
+}
+
/**
* topa_alloc() - allocate page-sized ToPA table
* @cpu: CPU on which to allocate.
@@ -812,6 +822,11 @@ static void pt_update_head(struct pt *pt)
struct pt_buffer *buf = perf_get_aux(&pt->handle);
u64 topa_idx, base, old;
+ if (buf->single) {
+ local_set(&buf->data_size, buf->output_off);
+ return;
+ }
+
/* offset of the first region in this table from the beginning of buf */
base = buf->cur->offset + buf->output_off;
@@ -913,18 +928,21 @@ static void pt_handle_status(struct pt *pt)
*/
static void pt_read_offset(struct pt_buffer *buf)
{
- u64 offset, base_topa;
+ u64 offset, base;
struct topa_page *tp;
- rdmsrl(MSR_IA32_RTIT_OUTPUT_BASE, base_topa);
- tp = phys_to_virt(base_topa);
- buf->cur = &tp->topa;
+ rdmsrl(MSR_IA32_RTIT_OUTPUT_BASE, base);
+ if (!buf->single) {
+ tp = phys_to_virt(base);
+ buf->cur = &tp->topa;
+ }
rdmsrl(MSR_IA32_RTIT_OUTPUT_MASK, offset);
/* offset within current output region */
buf->output_off = offset >> 32;
/* index of current output region within this table */
- buf->cur_idx = (offset & 0xffffff80) >> 7;
+ if (!buf->single)
+ buf->cur_idx = (offset & 0xffffff80) >> 7;
}
static struct topa_entry *
@@ -1040,6 +1058,9 @@ static int pt_buffer_reset_markers(struct pt_buffer *buf,
unsigned long head = local64_read(&buf->head);
unsigned long idx, npages, wakeup;
+ if (buf->single)
+ return 0;
+
/* can't stop in the middle of an output region */
if (buf->output_off + handle->size + 1 < pt_buffer_region_size(buf)) {
perf_aux_output_flag(handle, PERF_AUX_FLAG_TRUNCATED);
@@ -1121,13 +1142,17 @@ static void pt_buffer_reset_offsets(struct pt_buffer *buf, unsigned long head)
if (buf->snapshot)
head &= (buf->nr_pages << PAGE_SHIFT) - 1;
- pg = (head >> PAGE_SHIFT) & (buf->nr_pages - 1);
- te = pt_topa_entry_for_page(buf, pg);
+ if (!buf->single) {
+ pg = (head >> PAGE_SHIFT) & (buf->nr_pages - 1);
+ te = pt_topa_entry_for_page(buf, pg);
- cur_tp = topa_entry_to_page(te);
- buf->cur = &cur_tp->topa;
- buf->cur_idx = te - TOPA_ENTRY(buf->cur, 0);
- buf->output_off = head & (pt_buffer_region_size(buf) - 1);
+ cur_tp = topa_entry_to_page(te);
+ buf->cur = &cur_tp->topa;
+ buf->cur_idx = te - TOPA_ENTRY(buf->cur, 0);
+ buf->output_off = head & (pt_buffer_region_size(buf) - 1);
+ } else {
+ buf->output_off = head;
+ }
local64_set(&buf->head, head);
local_set(&buf->data_size, 0);
@@ -1141,6 +1166,9 @@ static void pt_buffer_fini_topa(struct pt_buffer *buf)
{
struct topa *topa, *iter;
+ if (buf->single)
+ return;
+
list_for_each_entry_safe(topa, iter, &buf->tables, list) {
/*
* right now, this is in free_aux() path only, so
@@ -1186,6 +1214,36 @@ static int pt_buffer_init_topa(struct pt_buffer *buf, int cpu,
return 0;
}
+static int pt_buffer_try_single(struct pt_buffer *buf, int nr_pages)
+{
+ struct page *p = virt_to_page(buf->data_pages[0]);
+ int ret = -ENOTSUPP, order = 0;
+
+ /*
+ * We can use single range output mode
+ * + in snapshot mode, where we don't need interrupts;
+ * + if the hardware supports it;
+ * + if the entire buffer is one contiguous allocation.
+ */
+ if (!buf->snapshot)
+ goto out;
+
+ if (!intel_pt_validate_hw_cap(PT_CAP_single_range_output))
+ goto out;
+
+ if (PagePrivate(p))
+ order = page_private(p);
+
+ if (1 << order != nr_pages)
+ goto out;
+
+ buf->single = true;
+ buf->nr_pages = nr_pages;
+ ret = 0;
+out:
+ return ret;
+}
+
/**
* pt_buffer_setup_aux() - set up topa tables for a PT buffer
* @cpu: Cpu on which to allocate, -1 means current.
@@ -1230,6 +1288,10 @@ pt_buffer_setup_aux(struct perf_event *event, void **pages,
INIT_LIST_HEAD(&buf->tables);
+ ret = pt_buffer_try_single(buf, nr_pages);
+ if (!ret)
+ return buf;
+
ret = pt_buffer_init_topa(buf, cpu, nr_pages, GFP_KERNEL);
if (ret) {
kfree(buf);
@@ -1396,8 +1458,7 @@ void intel_pt_interrupt(void)
return;
}
- pt_config_buffer(topa_to_page(buf->cur)->table, buf->cur_idx,
- buf->output_off);
+ pt_config_buffer(buf);
pt_config_start(event);
}
}
@@ -1461,8 +1522,7 @@ static void pt_event_start(struct perf_event *event, int mode)
WRITE_ONCE(pt->handle_nmi, 1);
hwc->state = 0;
- pt_config_buffer(topa_to_page(buf->cur)->table, buf->cur_idx,
- buf->output_off);
+ pt_config_buffer(buf);
pt_config(event);
return;
diff --git a/arch/x86/events/intel/pt.h b/arch/x86/events/intel/pt.h
index 1d2bb7572374..3f7818221b95 100644
--- a/arch/x86/events/intel/pt.h
+++ b/arch/x86/events/intel/pt.h
@@ -64,6 +64,7 @@ struct pt_pmu {
* @lost: if data was lost/truncated
* @head: logical write offset inside the buffer
* @snapshot: if this is for a snapshot/overwrite counter
+ * @single: use Single Range Output instead of ToPA
* @stop_pos: STOP topa entry index
* @intr_pos: INT topa entry index
* @stop_te: STOP topa entry pointer
@@ -80,6 +81,7 @@ struct pt_buffer {
local_t data_size;
local64_t head;
bool snapshot;
+ bool single;
long stop_pos, intr_pos;
struct topa_entry *stop_te, *intr_te;
void **data_pages;
--
2.23.0
next prev parent reply other threads:[~2019-10-22 9:58 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-10-22 9:58 [PATCH v2 0/4] perf: Add AUX data sampling Alexander Shishkin
2019-10-22 9:58 ` [PATCH v2 1/4] perf: Allow using AUX data in perf samples Alexander Shishkin
2019-10-24 13:52 ` Peter Zijlstra
2019-10-24 14:01 ` Peter Zijlstra
2019-10-25 12:52 ` Alexander Shishkin
2019-10-24 14:06 ` Peter Zijlstra
2019-10-25 12:21 ` Alexander Shishkin
2019-10-24 14:09 ` Peter Zijlstra
2019-10-24 14:12 ` Peter Zijlstra
2019-10-22 9:58 ` [PATCH v2 2/4] perf/x86/intel/pt: Factor out starting the trace Alexander Shishkin
2019-10-22 9:58 ` [PATCH v2 3/4] perf/x86/intel/pt: Add sampling support Alexander Shishkin
2019-10-22 9:58 ` Alexander Shishkin [this message]
2019-10-23 15:09 ` [PATCH v2 4/4] perf/x86/intel/pt: Opportunistically use single range output mode Alexander Shishkin
2019-10-24 13:56 ` Peter Zijlstra
2019-10-25 12:19 ` Alexander Shishkin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20191022095812.67071-5-alexander.shishkin@linux.intel.com \
--to=alexander.shishkin@linux.intel.com \
--cc=acme@redhat.com \
--cc=adrian.hunter@intel.com \
--cc=jolsa@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mark.rutland@arm.com \
--cc=mathieu.poirier@linaro.org \
--cc=mingo@redhat.com \
--cc=peterz@infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox