[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Qemu-devel] [PULL 41/46] rcu: reduce more than 7MB heap memory by mallo
From: |
Paolo Bonzini |
Subject: |
[Qemu-devel] [PULL 41/46] rcu: reduce more than 7MB heap memory by malloc_trim() |
Date: |
Wed, 20 Dec 2017 18:14:53 +0100 |
From: Yang Zhong <address@hidden>
Since there are some issues in memory alloc/free machenism
in glibc for little chunk memory, if Qemu frequently
alloc/free little chunk memory, the glibc doesn't alloc
little chunk memory from free list of glibc and still
allocate from OS, which make the heap size bigger and bigger.
This patch introduce malloc_trim(), which will free heap
memory when there is no rcu call during rcu thread loop.
malloc_trim() can be enabled/disabled by --enable-malloc-trim/
--disable-malloc-trim in the Qemu configure command. The
default malloc_trim() is enabled for libc.
Below are test results from smaps file.
(1)without patch
55f0783e1000-55f07992a000 rw-p 00000000 00:00 0 [heap]
Size: 21796 kB
Rss: 14260 kB
Pss: 14260 kB
(2)with patch
55cc5fadf000-55cc61008000 rw-p 00000000 00:00 0 [heap]
Size: 21668 kB
Rss: 6940 kB
Pss: 6940 kB
Signed-off-by: Yang Zhong <address@hidden>
Message-Id: <address@hidden>
Signed-off-by: Paolo Bonzini <address@hidden>
---
configure | 35 +++++++++++++++++++++++++++++++++++
util/rcu.c | 6 ++++++
2 files changed, 41 insertions(+)
diff --git a/configure b/configure
index 99ccc17..100309c 100755
--- a/configure
+++ b/configure
@@ -426,6 +426,7 @@ vxhs=""
supported_cpu="no"
supported_os="no"
bogus_os="no"
+malloc_trim=""
# parse CC options first
for opt do
@@ -1047,6 +1048,10 @@ for opt do
;;
--enable-tcg) tcg="yes"
;;
+ --disable-malloc-trim) malloc_trim="no"
+ ;;
+ --enable-malloc-trim) malloc_trim="yes"
+ ;;
--disable-spice) spice="no"
;;
--enable-spice) spice="yes"
@@ -1466,6 +1471,7 @@ Advanced options (experts only):
Default:trace-<pid>
--disable-slirp disable SLIRP userspace network connectivity
--enable-tcg-interpreter enable TCG with bytecode interpreter (TCI)
+ --enable-malloc-trim enable libc malloc_trim() for memory optimization
--oss-lib path to OSS library
--cpu=CPU Build for host CPU [$cpu]
--with-coroutine=BACKEND coroutine backend. Supported options:
@@ -3860,6 +3866,30 @@ if test "$tcmalloc" = "yes" && test "$jemalloc" = "yes"
; then
exit 1
fi
+# Even if malloc_trim() is available, these non-libc memory allocators
+# do not support it.
+if test "$tcmalloc" = "yes" || test "$jemalloc" = "yes" ; then
+ if test "$malloc_trim" = "yes" ; then
+ echo "Disabling malloc_trim with non-libc memory allocator"
+ fi
+ malloc_trim="no"
+fi
+
+#######################################
+# malloc_trim
+
+if test "$malloc_trim" != "no" ; then
+ cat > $TMPC << EOF
+#include <malloc.h>
+int main(void) { malloc_trim(0); return 0; }
+EOF
+ if compile_prog "" "" ; then
+ malloc_trim="yes"
+ else
+ malloc_trim="no"
+ fi
+fi
+
##########################################
# tcmalloc probe
@@ -5505,6 +5535,7 @@ if test "$tcg" = "yes" ; then
echo "TCG debug enabled $debug_tcg"
echo "TCG interpreter $tcg_interpreter"
fi
+echo "malloc trim support $malloc_trim"
echo "RDMA support $rdma"
echo "fdt support $fdt"
echo "preadv support $preadv"
@@ -6015,6 +6046,10 @@ if test "$opengl" = "yes" ; then
fi
fi
+if test "$malloc_trim" = "yes" ; then
+ echo "CONFIG_MALLOC_TRIM=y" >> $config_host_mak
+fi
+
if test "$avx2_opt" = "yes" ; then
echo "CONFIG_AVX2_OPT=y" >> $config_host_mak
fi
diff --git a/util/rcu.c b/util/rcu.c
index ca5a63e..f4d09c8 100644
--- a/util/rcu.c
+++ b/util/rcu.c
@@ -32,6 +32,9 @@
#include "qemu/atomic.h"
#include "qemu/thread.h"
#include "qemu/main-loop.h"
+#if defined(CONFIG_MALLOC_TRIM)
+#include <malloc.h>
+#endif
/*
* Global grace period counter. Bit 0 is always one in rcu_gp_ctr.
@@ -246,6 +249,9 @@ static void *call_rcu_thread(void *opaque)
qemu_event_reset(&rcu_call_ready_event);
n = atomic_read(&rcu_call_count);
if (n == 0) {
+#if defined(CONFIG_MALLOC_TRIM)
+ malloc_trim(4 * 1024 * 1024);
+#endif
qemu_event_wait(&rcu_call_ready_event);
}
}
--
1.8.3.1
- [Qemu-devel] [PULL 33/46] scsi: replace hex constants with #defines, (continued)
- [Qemu-devel] [PULL 33/46] scsi: replace hex constants with #defines, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 32/46] scsi: provide general-purpose functions to manage sense data, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 37/46] i8259: generalize statistics into common code, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 35/46] i8259: convert DPRINTFs into trace, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 38/46] kvm-i8259: support "info pic" and "info irq", Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 36/46] i8259: use DEBUG_IRQ_COUNT always, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 34/46] Remove legacy -no-kvm-pit option, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 39/46] i8259: move TYPE_INTERRUPT_STATS_PROVIDER upper, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 40/46] checkpatch: volatile with a comment or sig_atomic_t is okay, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 43/46] test: add some chardev mux event tests, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 41/46] rcu: reduce more than 7MB heap memory by malloc_trim(),
Paolo Bonzini <=
- [Qemu-devel] [PULL 42/46] chardev: fix backend events regression with mux chardev, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 45/46] blockdev: convert qemu-nbd server to QIONetListener, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 44/46] blockdev: convert internal NBD server to QIONetListener, Paolo Bonzini, 2017/12/20
- [Qemu-devel] [PULL 46/46] chardev: convert the socket server to QIONetListener, Paolo Bonzini, 2017/12/20
- Re: [Qemu-devel] [PULL 00/46] First batch of misc patches for QEMU 2.12, Peter Maydell, 2017/12/20