From: Arnaldo Carvalho de Melo <acme@kernel.org>
To: Ingo Molnar <mingo@kernel.org>
Cc: linux-kernel@vger.kernel.org,
Stefano Sanfilippo <ssanfilippo@chromium.org>,
Ross McIlroy <rmcilroy@chromium.org>,
Anton Blanchard <anton@ozlabs.org>, Jiri Olsa <jolsa@redhat.com>,
Namhyung Kim <namhyung@kernel.org>,
Peter Zijlstra <peterz@infradead.org>,
Arnaldo Carvalho de Melo <acme@redhat.com>
Subject: [PATCH 22/37] perf jit: Add unwinding support
Date: Mon, 24 Oct 2016 13:20:42 -0300 [thread overview]
Message-ID: <1477326057-24080-23-git-send-email-acme@kernel.org> (raw)
In-Reply-To: <1477326057-24080-1-git-send-email-acme@kernel.org>
From: Stefano Sanfilippo <ssanfilippo@chromium.org>
This record is intended to provide unwinding information in the
eh_frame format. This is required to unwind JITed code which
does not maintain the frame pointer register during function calls.
The eh_frame unwinding information can be emitted by V8 / Chromium
when the --perf_prof_unwinding_info is passed.
A record of type jr_code_unwinding_info comes before the jr_code_load
it referred to and contains both the .eh_frame and .eh_frame_hdr.
The fields in the header have the following meaning:
* unwinding_size: size of the eh_frame and eh_frame_hdr, necessary
for distinguishing the content from the padding.
* eh_frame_hdr_size: as the name says.
* mapped_size: size of the payload that was in memory at runtime.
typically unwinding_size if the .eh_frame_hdr and .eh_frame were
mapped, or 0 if they weren't. It should always be the former case,
since the .eh_frame is guaranteed to be mapped in memory. However,
certain JITs might want to inject an .eh_frame_hdr with an empty LUT
to trigger fp-based unwinding fallback in libunwind. The only part
of the .eh_frame_hdr that libunwind reads from remote memory is the
LUT, and since there is none, mapping the unwinding info in memory
is not necessary, and 0 in this field signifies that it wasn't.
This practical hack allows to save bytes in code memory for those
JIT compilers that might or might not maintain a valid frame pointer.
The payload that follows is assumed to contain first the .eh_frame and
then the .eh_header_hdr, with no padding between the two.
Signed-off-by: Stefano Sanfilippo <ssanfilippo@chromium.org>
Signed-off-by: Ross McIlroy <rmcilroy@chromium.org>
Reviewed-by: Stephane Eranian <eranian@google.com>
Cc: Anton Blanchard <anton@ozlabs.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: http://lkml.kernel.org/r/1476356383-30100-7-git-send-email-eranian@google.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
---
tools/perf/util/jitdump.c | 57 ++++++++++++++++++++++++++++++++++++++++++++---
tools/perf/util/jitdump.h | 12 ++++++++++
2 files changed, 66 insertions(+), 3 deletions(-)
diff --git a/tools/perf/util/jitdump.c b/tools/perf/util/jitdump.c
index 75b66bbf8429..9bae66cc78f2 100644
--- a/tools/perf/util/jitdump.c
+++ b/tools/perf/util/jitdump.c
@@ -37,6 +37,10 @@ struct jit_buf_desc {
bool needs_bswap; /* handles cross-endianess */
bool use_arch_timestamp;
void *debug_data;
+ void *unwinding_data;
+ uint64_t unwinding_size;
+ uint64_t unwinding_mapped_size;
+ uint64_t eh_frame_hdr_size;
size_t nr_debug_entries;
uint32_t code_load_count;
u64 bytes_written;
@@ -295,6 +299,13 @@ jit_get_next_entry(struct jit_buf_desc *jd)
}
}
break;
+ case JIT_CODE_UNWINDING_INFO:
+ if (jd->needs_bswap) {
+ jr->unwinding.unwinding_size = bswap_64(jr->unwinding.unwinding_size);
+ jr->unwinding.eh_frame_hdr_size = bswap_64(jr->unwinding.eh_frame_hdr_size);
+ jr->unwinding.mapped_size = bswap_64(jr->unwinding.mapped_size);
+ }
+ break;
case JIT_CODE_CLOSE:
break;
case JIT_CODE_LOAD:
@@ -370,7 +381,7 @@ static int jit_repipe_code_load(struct jit_buf_desc *jd, union jr_entry *jr)
u16 idr_size;
const char *sym;
uint32_t count;
- int ret, csize;
+ int ret, csize, usize;
pid_t pid, tid;
struct {
u32 pid, tid;
@@ -380,6 +391,7 @@ static int jit_repipe_code_load(struct jit_buf_desc *jd, union jr_entry *jr)
pid = jr->load.pid;
tid = jr->load.tid;
csize = jr->load.code_size;
+ usize = jd->unwinding_mapped_size;
addr = jr->load.code_addr;
sym = (void *)((unsigned long)jr + sizeof(jr->load));
code = (unsigned long)jr + jr->load.p.total_size - csize;
@@ -408,6 +420,14 @@ static int jit_repipe_code_load(struct jit_buf_desc *jd, union jr_entry *jr)
jd->nr_debug_entries = 0;
}
+ if (jd->unwinding_data && jd->eh_frame_hdr_size) {
+ free(jd->unwinding_data);
+ jd->unwinding_data = NULL;
+ jd->eh_frame_hdr_size = 0;
+ jd->unwinding_mapped_size = 0;
+ jd->unwinding_size = 0;
+ }
+
if (ret) {
free(event);
return -1;
@@ -422,7 +442,7 @@ static int jit_repipe_code_load(struct jit_buf_desc *jd, union jr_entry *jr)
event->mmap2.pgoff = GEN_ELF_TEXT_OFFSET;
event->mmap2.start = addr;
- event->mmap2.len = csize;
+ event->mmap2.len = usize ? ALIGN_8(csize) + usize : csize;
event->mmap2.pid = pid;
event->mmap2.tid = tid;
event->mmap2.ino = st.st_ino;
@@ -473,6 +493,7 @@ static int jit_repipe_code_move(struct jit_buf_desc *jd, union jr_entry *jr)
char *filename;
size_t size;
struct stat st;
+ int usize;
u16 idr_size;
int ret;
pid_t pid, tid;
@@ -483,6 +504,7 @@ static int jit_repipe_code_move(struct jit_buf_desc *jd, union jr_entry *jr)
pid = jr->move.pid;
tid = jr->move.tid;
+ usize = jd->unwinding_mapped_size;
idr_size = jd->machine->id_hdr_size;
/*
@@ -511,7 +533,8 @@ static int jit_repipe_code_move(struct jit_buf_desc *jd, union jr_entry *jr)
(sizeof(event->mmap2.filename) - size) + idr_size);
event->mmap2.pgoff = GEN_ELF_TEXT_OFFSET;
event->mmap2.start = jr->move.new_code_addr;
- event->mmap2.len = jr->move.code_size;
+ event->mmap2.len = usize ? ALIGN_8(jr->move.code_size) + usize
+ : jr->move.code_size;
event->mmap2.pid = pid;
event->mmap2.tid = tid;
event->mmap2.ino = st.st_ino;
@@ -578,6 +601,31 @@ static int jit_repipe_debug_info(struct jit_buf_desc *jd, union jr_entry *jr)
}
static int
+jit_repipe_unwinding_info(struct jit_buf_desc *jd, union jr_entry *jr)
+{
+ void *unwinding_data;
+ uint32_t unwinding_data_size;
+
+ if (!(jd && jr))
+ return -1;
+
+ unwinding_data_size = jr->prefix.total_size - sizeof(jr->unwinding);
+ unwinding_data = malloc(unwinding_data_size);
+ if (!unwinding_data)
+ return -1;
+
+ memcpy(unwinding_data, &jr->unwinding.unwinding_data,
+ unwinding_data_size);
+
+ jd->eh_frame_hdr_size = jr->unwinding.eh_frame_hdr_size;
+ jd->unwinding_size = jr->unwinding.unwinding_size;
+ jd->unwinding_mapped_size = jr->unwinding.mapped_size;
+ jd->unwinding_data = unwinding_data;
+
+ return 0;
+}
+
+static int
jit_process_dump(struct jit_buf_desc *jd)
{
union jr_entry *jr;
@@ -594,6 +642,9 @@ jit_process_dump(struct jit_buf_desc *jd)
case JIT_CODE_DEBUG_INFO:
ret = jit_repipe_debug_info(jd, jr);
break;
+ case JIT_CODE_UNWINDING_INFO:
+ ret = jit_repipe_unwinding_info(jd, jr);
+ break;
default:
ret = 0;
continue;
diff --git a/tools/perf/util/jitdump.h b/tools/perf/util/jitdump.h
index bcacd20d0c1c..c6b9b67f43bf 100644
--- a/tools/perf/util/jitdump.h
+++ b/tools/perf/util/jitdump.h
@@ -19,6 +19,7 @@
#define JITHEADER_MAGIC_SW 0x4454694A
#define PADDING_8ALIGNED(x) ((((x) + 7) & 7) ^ 7)
+#define ALIGN_8(x) (((x) + 7) & (~7))
#define JITHEADER_VERSION 1
@@ -48,6 +49,7 @@ enum jit_record_type {
JIT_CODE_MOVE = 1,
JIT_CODE_DEBUG_INFO = 2,
JIT_CODE_CLOSE = 3,
+ JIT_CODE_UNWINDING_INFO = 4,
JIT_CODE_MAX,
};
@@ -101,12 +103,22 @@ struct jr_code_debug_info {
struct debug_entry entries[0];
};
+struct jr_code_unwinding_info {
+ struct jr_prefix p;
+
+ uint64_t unwinding_size;
+ uint64_t eh_frame_hdr_size;
+ uint64_t mapped_size;
+ const char unwinding_data[0];
+};
+
union jr_entry {
struct jr_code_debug_info info;
struct jr_code_close close;
struct jr_code_load load;
struct jr_code_move move;
struct jr_prefix prefix;
+ struct jr_code_unwinding_info unwinding;
};
static inline struct debug_entry *
--
2.7.4
next prev parent reply other threads:[~2016-10-24 16:22 UTC|newest]
Thread overview: 42+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-10-24 16:20 [GIT PULL 00/37] perf/core improvements and fixes Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 01/37] perf intel-pt/bts: Tidy instruction buffer size usage Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 02/37] perf intel-pt/bts: Report instruction bytes and length in sample Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 03/37] perf script: Support insn and insnlen Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 04/37] perf tools: Sync copy of x86's syscall table Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 05/37] tools lib traceevent: Add install_headers target Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 06/37] tools lib traceevent: Add do_install_mkdir Makefile function Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 07/37] tools lib traceevent: Rename LIB_FILE to LIB_TARGET Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 08/37] tools lib traceevent: Add version for traceevent shared object Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 09/37] tools lib: Add for_each_clear_bit macro Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 10/37] perf report: Move captured info to generic header info Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 11/37] perf header: Display missing features Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 12/37] perf header: Display feature name on write failure Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 13/37] perf record: Improve documentation of event parameters Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 14/37] perf tools: Implement branch_type event parameter Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 15/37] perf jit: Avoid returning garbage for a ret variable Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 16/37] perf jit: Add NT_GNU_BUILD_ID definition for older distros Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 17/37] perf jit: Improve error messages from JVMTI Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 18/37] perf jit: Enable jitdump support without dwarf Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 19/37] perf jit: Remove unecessary padding in jitdump file Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 20/37] perf jit: Make perf skip unknown records Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 21/37] perf jit: Do not assume pgoff is zero Arnaldo Carvalho de Melo
2016-10-24 16:20 ` Arnaldo Carvalho de Melo [this message]
2016-10-24 16:20 ` [PATCH 23/37] perf jit: Generate .eh_frame/.eh_frame_hdr in DSO Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 24/37] perf jit: Check JITHEADER_VERSION Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 25/37] perf jit: Add jitdump format specification document Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 26/37] perf pmu: Only print Using CPUID message once Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 27/37] perf tools: Fix typo "No enough" to "Not enough" Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 28/37] perf hists browser: Dynamically change verbosity level Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 29/37] perf trace: Implement --delay Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 30/37] perf bench mem: Move boilerplate memory allocation to the infrastructure Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 31/37] perf tools: Normalize sq_quote_argv() error reporting Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 32/37] perf tools: Use normal error reporting when processing PERF_RECORD_READ events Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 33/37] perf bench futex: Cache align the worker struct Arnaldo Carvalho de Melo
2016-10-24 18:50 ` Davidlohr Bueso
2016-10-24 18:53 ` Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 34/37] perf trace: Remove thread_trace->exit_time Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 35/37] perf trace: Use the syscall raw_syscalls:sys_enter timestamp Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 36/37] perf list: Make vendor event matching case insensitive Arnaldo Carvalho de Melo
2016-10-24 16:20 ` [PATCH 37/37] perf coresight: Removing miscellaneous debug output Arnaldo Carvalho de Melo
2016-10-24 16:20 ` Arnaldo Carvalho de Melo
2016-10-24 18:44 ` [GIT PULL 00/37] perf/core improvements and fixes Ingo Molnar
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=1477326057-24080-23-git-send-email-acme@kernel.org \
--to=acme@kernel.org \
--cc=acme@redhat.com \
--cc=anton@ozlabs.org \
--cc=jolsa@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@kernel.org \
--cc=namhyung@kernel.org \
--cc=peterz@infradead.org \
--cc=rmcilroy@chromium.org \
--cc=ssanfilippo@chromium.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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.