qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: Stefan Hajnoczi <stefanha@redhat.com>
To: qemu-devel@nongnu.org
Cc: Paolo Bonzini <pbonzini@redhat.com>,
	"Dr. David Alan Gilbert" <dgilbert@redhat.com>,
	Richard Henderson <rth@twiddle.net>,
	Peter Crosthwaite <crosthwaite.peter@gmail.com>,
	Stefan Hajnoczi <stefanha@redhat.com>,
	Juan Quintela <quintela@redhat.com>,
	Markus Armbruster <armbru@redhat.com>,
	Michael Roth <mdroth@linux.vnet.ibm.com>,
	Peter Maydell <peter.maydell@linaro.org>,
	"Emilio G. Cota" <cota@braap.org>
Subject: [Qemu-devel] [PULL 5/6] trace: enable tracing of TCG atomics
Date: Wed, 27 Jun 2018 13:58:46 +0100	[thread overview]
Message-ID: <20180627125847.5413-6-stefanha@redhat.com> (raw)
In-Reply-To: <20180627125847.5413-1-stefanha@redhat.com>

From: "Emilio G. Cota" <cota@braap.org>

We do not trace guest atomic accesses. Fix it.

Tested with a modified atomic_add-bench so that it executes
a deterministic number of instructions, i.e. fixed seeding,
no threading and fixed number of loop iterations instead
of running for a certain time.

Before:
- With parallel_cpus = false (no clone syscall so it is never set to true):
  220070 memory accesses
- With parallel_cpus = true (hard-coded):
  212105 memory accesses <-- we're not tracing the atomics!

After:
  220070 memory accesses regardless of parallel_cpus.

Signed-off-by: Emilio G. Cota <cota@braap.org>
Message-id: 1527028012-21888-6-git-send-email-cota@braap.org
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
 accel/tcg/atomic_template.h | 87 ++++++++++++++++++++++++++++++++++---
 1 file changed, 81 insertions(+), 6 deletions(-)

diff --git a/accel/tcg/atomic_template.h b/accel/tcg/atomic_template.h
index 3f41ef2782..d751bcba48 100644
--- a/accel/tcg/atomic_template.h
+++ b/accel/tcg/atomic_template.h
@@ -18,30 +18,37 @@
  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
  */
 
+#include "trace/mem.h"
+
 #if DATA_SIZE == 16
 # define SUFFIX     o
 # define DATA_TYPE  Int128
 # define BSWAP      bswap128
+# define SHIFT      4
 #elif DATA_SIZE == 8
 # define SUFFIX     q
 # define DATA_TYPE  uint64_t
 # define SDATA_TYPE int64_t
 # define BSWAP      bswap64
+# define SHIFT      3
 #elif DATA_SIZE == 4
 # define SUFFIX     l
 # define DATA_TYPE  uint32_t
 # define SDATA_TYPE int32_t
 # define BSWAP      bswap32
+# define SHIFT      2
 #elif DATA_SIZE == 2
 # define SUFFIX     w
 # define DATA_TYPE  uint16_t
 # define SDATA_TYPE int16_t
 # define BSWAP      bswap16
+# define SHIFT      1
 #elif DATA_SIZE == 1
 # define SUFFIX     b
 # define DATA_TYPE  uint8_t
 # define SDATA_TYPE int8_t
 # define BSWAP
+# define SHIFT      0
 #else
 # error unsupported data size
 #endif
@@ -52,14 +59,37 @@
 # define ABI_TYPE  uint32_t
 #endif
 
+#define ATOMIC_TRACE_RMW do {                                           \
+        uint8_t info = glue(trace_mem_build_info_no_se, MEND)(SHIFT, false); \
+                                                                        \
+        trace_guest_mem_before_exec(ENV_GET_CPU(env), addr, info);      \
+        trace_guest_mem_before_exec(ENV_GET_CPU(env), addr,             \
+                                    info | TRACE_MEM_ST);               \
+    } while (0)
+
+#define ATOMIC_TRACE_LD do {                                            \
+        uint8_t info = glue(trace_mem_build_info_no_se, MEND)(SHIFT, false); \
+                                                                        \
+        trace_guest_mem_before_exec(ENV_GET_CPU(env), addr, info);      \
+    } while (0)
+
+# define ATOMIC_TRACE_ST do {                                           \
+        uint8_t info = glue(trace_mem_build_info_no_se, MEND)(SHIFT, true); \
+                                                                        \
+        trace_guest_mem_before_exec(ENV_GET_CPU(env), addr, info);      \
+    } while (0)
+
 /* Define host-endian atomic operations.  Note that END is used within
    the ATOMIC_NAME macro, and redefined below.  */
 #if DATA_SIZE == 1
 # define END
+# define MEND _be /* either le or be would be fine */
 #elif defined(HOST_WORDS_BIGENDIAN)
 # define END  _be
+# define MEND _be
 #else
 # define END  _le
+# define MEND _le
 #endif
 
 ABI_TYPE ATOMIC_NAME(cmpxchg)(CPUArchState *env, target_ulong addr,
@@ -67,7 +97,10 @@ ABI_TYPE ATOMIC_NAME(cmpxchg)(CPUArchState *env, target_ulong addr,
 {
     ATOMIC_MMU_DECLS;
     DATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;
-    DATA_TYPE ret = atomic_cmpxchg__nocheck(haddr, cmpv, newv);
+    DATA_TYPE ret;
+
+    ATOMIC_TRACE_RMW;
+    ret = atomic_cmpxchg__nocheck(haddr, cmpv, newv);
     ATOMIC_MMU_CLEANUP;
     return ret;
 }
@@ -77,6 +110,8 @@ ABI_TYPE ATOMIC_NAME(ld)(CPUArchState *env, target_ulong addr EXTRA_ARGS)
 {
     ATOMIC_MMU_DECLS;
     DATA_TYPE val, *haddr = ATOMIC_MMU_LOOKUP;
+
+    ATOMIC_TRACE_LD;
     __atomic_load(haddr, &val, __ATOMIC_RELAXED);
     ATOMIC_MMU_CLEANUP;
     return val;
@@ -87,6 +122,8 @@ void ATOMIC_NAME(st)(CPUArchState *env, target_ulong addr,
 {
     ATOMIC_MMU_DECLS;
     DATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;
+
+    ATOMIC_TRACE_ST;
     __atomic_store(haddr, &val, __ATOMIC_RELAXED);
     ATOMIC_MMU_CLEANUP;
 }
@@ -96,7 +133,10 @@ ABI_TYPE ATOMIC_NAME(xchg)(CPUArchState *env, target_ulong addr,
 {
     ATOMIC_MMU_DECLS;
     DATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;
-    DATA_TYPE ret = atomic_xchg__nocheck(haddr, val);
+    DATA_TYPE ret;
+
+    ATOMIC_TRACE_RMW;
+    ret = atomic_xchg__nocheck(haddr, val);
     ATOMIC_MMU_CLEANUP;
     return ret;
 }
@@ -107,7 +147,10 @@ ABI_TYPE ATOMIC_NAME(X)(CPUArchState *env, target_ulong addr,       \
 {                                                                   \
     ATOMIC_MMU_DECLS;                                               \
     DATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;                           \
-    DATA_TYPE ret = atomic_##X(haddr, val);                         \
+    DATA_TYPE ret;                                                  \
+                                                                    \
+    ATOMIC_TRACE_RMW;                                               \
+    ret = atomic_##X(haddr, val);                                   \
     ATOMIC_MMU_CLEANUP;                                             \
     return ret;                                                     \
 }
@@ -126,6 +169,9 @@ GEN_ATOMIC_HELPER(xor_fetch)
 /* These helpers are, as a whole, full barriers.  Within the helper,
  * the leading barrier is explicit and the trailing barrier is within
  * cmpxchg primitive.
+ *
+ * Trace this load + RMW loop as a single RMW op. This way, regardless
+ * of CF_PARALLEL's value, we'll trace just a read and a write.
  */
 #define GEN_ATOMIC_HELPER_FN(X, FN, XDATA_TYPE, RET)                \
 ABI_TYPE ATOMIC_NAME(X)(CPUArchState *env, target_ulong addr,       \
@@ -134,6 +180,8 @@ ABI_TYPE ATOMIC_NAME(X)(CPUArchState *env, target_ulong addr,       \
     ATOMIC_MMU_DECLS;                                               \
     XDATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;                          \
     XDATA_TYPE cmp, old, new, val = xval;                           \
+                                                                    \
+    ATOMIC_TRACE_RMW;                                               \
     smp_mb();                                                       \
     cmp = atomic_read__nocheck(haddr);                              \
     do {                                                            \
@@ -158,6 +206,7 @@ GEN_ATOMIC_HELPER_FN(umax_fetch, MAX,  DATA_TYPE, new)
 #endif /* DATA SIZE >= 16 */
 
 #undef END
+#undef MEND
 
 #if DATA_SIZE > 1
 
@@ -165,8 +214,10 @@ GEN_ATOMIC_HELPER_FN(umax_fetch, MAX,  DATA_TYPE, new)
    within the ATOMIC_NAME macro.  */
 #ifdef HOST_WORDS_BIGENDIAN
 # define END  _le
+# define MEND _le
 #else
 # define END  _be
+# define MEND _be
 #endif
 
 ABI_TYPE ATOMIC_NAME(cmpxchg)(CPUArchState *env, target_ulong addr,
@@ -174,7 +225,10 @@ ABI_TYPE ATOMIC_NAME(cmpxchg)(CPUArchState *env, target_ulong addr,
 {
     ATOMIC_MMU_DECLS;
     DATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;
-    DATA_TYPE ret = atomic_cmpxchg__nocheck(haddr, BSWAP(cmpv), BSWAP(newv));
+    DATA_TYPE ret;
+
+    ATOMIC_TRACE_RMW;
+    ret = atomic_cmpxchg__nocheck(haddr, BSWAP(cmpv), BSWAP(newv));
     ATOMIC_MMU_CLEANUP;
     return BSWAP(ret);
 }
@@ -184,6 +238,8 @@ ABI_TYPE ATOMIC_NAME(ld)(CPUArchState *env, target_ulong addr EXTRA_ARGS)
 {
     ATOMIC_MMU_DECLS;
     DATA_TYPE val, *haddr = ATOMIC_MMU_LOOKUP;
+
+    ATOMIC_TRACE_LD;
     __atomic_load(haddr, &val, __ATOMIC_RELAXED);
     ATOMIC_MMU_CLEANUP;
     return BSWAP(val);
@@ -194,6 +250,8 @@ void ATOMIC_NAME(st)(CPUArchState *env, target_ulong addr,
 {
     ATOMIC_MMU_DECLS;
     DATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;
+
+    ATOMIC_TRACE_ST;
     val = BSWAP(val);
     __atomic_store(haddr, &val, __ATOMIC_RELAXED);
     ATOMIC_MMU_CLEANUP;
@@ -204,7 +262,10 @@ ABI_TYPE ATOMIC_NAME(xchg)(CPUArchState *env, target_ulong addr,
 {
     ATOMIC_MMU_DECLS;
     DATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;
-    ABI_TYPE ret = atomic_xchg__nocheck(haddr, BSWAP(val));
+    ABI_TYPE ret;
+
+    ATOMIC_TRACE_RMW;
+    ret = atomic_xchg__nocheck(haddr, BSWAP(val));
     ATOMIC_MMU_CLEANUP;
     return BSWAP(ret);
 }
@@ -215,7 +276,10 @@ ABI_TYPE ATOMIC_NAME(X)(CPUArchState *env, target_ulong addr,       \
 {                                                                   \
     ATOMIC_MMU_DECLS;                                               \
     DATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;                           \
-    DATA_TYPE ret = atomic_##X(haddr, BSWAP(val));                  \
+    DATA_TYPE ret;                                                  \
+                                                                    \
+    ATOMIC_TRACE_RMW;                                               \
+    ret = atomic_##X(haddr, BSWAP(val));                            \
     ATOMIC_MMU_CLEANUP;                                             \
     return BSWAP(ret);                                              \
 }
@@ -232,6 +296,9 @@ GEN_ATOMIC_HELPER(xor_fetch)
 /* These helpers are, as a whole, full barriers.  Within the helper,
  * the leading barrier is explicit and the trailing barrier is within
  * cmpxchg primitive.
+ *
+ * Trace this load + RMW loop as a single RMW op. This way, regardless
+ * of CF_PARALLEL's value, we'll trace just a read and a write.
  */
 #define GEN_ATOMIC_HELPER_FN(X, FN, XDATA_TYPE, RET)                \
 ABI_TYPE ATOMIC_NAME(X)(CPUArchState *env, target_ulong addr,       \
@@ -240,6 +307,8 @@ ABI_TYPE ATOMIC_NAME(X)(CPUArchState *env, target_ulong addr,       \
     ATOMIC_MMU_DECLS;                                               \
     XDATA_TYPE *haddr = ATOMIC_MMU_LOOKUP;                          \
     XDATA_TYPE ldo, ldn, old, new, val = xval;                      \
+                                                                    \
+    ATOMIC_TRACE_RMW;                                               \
     smp_mb();                                                       \
     ldn = atomic_read__nocheck(haddr);                              \
     do {                                                            \
@@ -271,11 +340,17 @@ GEN_ATOMIC_HELPER_FN(add_fetch, ADD, DATA_TYPE, new)
 #endif /* DATA_SIZE >= 16 */
 
 #undef END
+#undef MEND
 #endif /* DATA_SIZE > 1 */
 
+#undef ATOMIC_TRACE_ST
+#undef ATOMIC_TRACE_LD
+#undef ATOMIC_TRACE_RMW
+
 #undef BSWAP
 #undef ABI_TYPE
 #undef DATA_TYPE
 #undef SDATA_TYPE
 #undef SUFFIX
 #undef DATA_SIZE
+#undef SHIFT
-- 
2.17.1

  parent reply	other threads:[~2018-06-27 12:59 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-06-27 12:58 [Qemu-devel] [PULL 0/6] Tracing patches Stefan Hajnoczi
2018-06-27 12:58 ` [Qemu-devel] [PULL 1/6] trace: fix misreporting of TCG access sizes for user-space Stefan Hajnoczi
2018-06-27 12:58 ` [Qemu-devel] [PULL 2/6] trace: simplify trace_mem functions Stefan Hajnoczi
2018-06-27 12:58 ` [Qemu-devel] [PULL 3/6] trace: expand mem_info:size_shift to 3 bits Stefan Hajnoczi
2018-06-27 12:58 ` [Qemu-devel] [PULL 4/6] trace: add trace_mem_build_info_no_se_be/le Stefan Hajnoczi
2018-06-27 12:58 ` Stefan Hajnoczi [this message]
2018-06-27 12:58 ` [Qemu-devel] [PULL 6/6] trace: forbid floating point types Stefan Hajnoczi
2018-06-28 13:26 ` [Qemu-devel] [PULL 0/6] Tracing patches Peter Maydell

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=20180627125847.5413-6-stefanha@redhat.com \
    --to=stefanha@redhat.com \
    --cc=armbru@redhat.com \
    --cc=cota@braap.org \
    --cc=crosthwaite.peter@gmail.com \
    --cc=dgilbert@redhat.com \
    --cc=mdroth@linux.vnet.ibm.com \
    --cc=pbonzini@redhat.com \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-devel@nongnu.org \
    --cc=quintela@redhat.com \
    --cc=rth@twiddle.net \
    /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;
as well as URLs for NNTP newsgroup(s).