On 01/31/2013 02:11 PM, xtu4 wrote:
[SEQ_FILE] Avoid high order memory allocating with kmalloc
when read large seq file
currently, when dumpstate access /proc/xxx/binder , this binder
include lots of info,
it will use seq_read in kernel, in this function, it will trigger high
order memory alloc,
when read binder info or other large file, this will cause memory
presure when system
don't have contious high order memory, it will lead to high kswap
workload to reclaim the
page. so change kmalloc to vmalloc, it can avoid contiously high order
memory allocating.
[ 4356.532357] dumpstate: page allocation failure: order:4, mode:0x40d0
[ 4356.532400] Pid: 18256, comm: dumpstate Tainted: G C
3.0.34-141128-g4be7088 #1
[ 4356.532416] Call Trace:
[ 4356.532443] [<c185c836>] ? printk+0x1d/0x1f
[ 4356.532467] [<c12cde1f>] warn_alloc_failed+0xbf/0xf0
[ 4356.532491] [<c12d0aba>] __alloc_pages_nodemask+0x4ba/0x6a0
[ 4356.532521] [<c12d0d1c>] __get_free_pages+0x1c/0x30
[ 4356.532541] [<c12f51e1>] kmalloc_order_trace+0x21/0xd0
[ 4356.532561] [<c131d0f7>] ? seq_read+0x137/0x390
[ 4356.532579] [<c12f549a>] __kmalloc+0x20a/0x230
[ 4356.532596] [<c131d0f7>] ? seq_read+0x137/0x390
[ 4356.532616] [<c12d343c>] ? put_page+0x2c/0x40
[ 4356.532634] [<c12f4a7d>] ? kfree+0xcd/0x160
[ 4356.532655] [<c18656cd>] ? mutex_unlock+0xd/0x10
[ 4356.532675] [<c131d109>] seq_read+0x149/0x390
[ 4356.532697] [<c130062c>] vfs_read+0x8c/0x160
[ 4356.532716] [<c131cfc0>] ? seq_lseek+0x180/0x180
[ 4356.532735] [<c130073d>] sys_read+0x3d/0x70
[ 4356.532755] [<c1866e91>] syscall_call+0x7/0xb
[ 4356.532777] [<c1860000>] ? log_dir_items+0x33d/0x40c
the m->size is very huge
<3>[ 1185.457656, 1] xiaobing >> seq_read: m->size 8192
<3>[ 1185.463462, 1] xiaobing >> seq_read: m->size 16384
<3>[ 1185.470472, 1] xiaobing >> seq_read: m->size 32768
<3>[ 1185.481201, 0] xiaobing >> seq_read: m->size 8192
<3>[ 1185.488071, 0] xiaobing >> seq_read: m->size 16384
<3>[ 1185.495892, 0] xiaobing >> seq_read: m->size 32768
<3>[ 1185.504841, 0] xiaobing >> seq_read: m->size 65536
<3>[ 1185.517180, 0] xiaobing >> seq_read: m->size 131072
<3>[ 1185.536286, 0] xiaobing >> seq_read: m->size 262144
some times even more then 262144 byte
Signed-off-by: xiaobing tu <xiaobing...@intel.com>
Change-Id: I892c97d02cf25e59b23c9bc68dff754ea01c1d56
---
fs/seq_file.c | 22 +++++++++++++++++-----
1 files changed, 17 insertions(+), 5 deletions(-)
diff --git a/fs/seq_file.c b/fs/seq_file.c
index dba43c3..19df826 100644
--- a/fs/seq_file.c
+++ b/fs/seq_file.c
@@ -12,7 +12,7 @@
#include <asm/uaccess.h>
#include <asm/page.h>
-
+#include <linux/mm.h>
/**
* seq_open - initialize sequential file
* @file: file we initialize
@@ -116,7 +116,13 @@ static int traverse(struct seq_file *m, loff_t
offset)
Eoverflow:
m->op->stop(m, p);
kfree(m->buf);
- m->buf = kmalloc(m->size <<= 1, GFP_KERNEL);
+
+ is_vmalloc_addr(m->buf) ? vfree(m->buf) : kfree(m->buf);
+ m->size <<= 1;
+ if (m->size <= (2 * PAGE_SIZE))
+ m->buf = kmalloc(m->size, GFP_KERNEL);
+ else
+ m->buf = vmalloc(m->size);
return !m->buf ? -ENOMEM : -EAGAIN;
}
@@ -209,8 +215,14 @@ ssize_t seq_read(struct file *file, char __user
*buf, size_t size, loff_t *ppos)
if (m->count < m->size)
goto Fill;
m->op->stop(m, p);
- kfree(m->buf);
- m->buf = kmalloc(m->size <<= 1, GFP_KERNEL);
+ is_vmalloc_addr(m->buf) ? vfree(m->buf) : kfree(m->buf);
+ m->size <<= 1;
+ if (m->size > 2 * PAGE_SIZE)
+ m->buf = vmalloc(m->size);
+ else
+ m->buf = kmalloc(m->size, GFP_KERNEL);
+
+
if (!m->buf)
goto Enomem;
m->count = 0;
@@ -325,7 +337,7 @@ EXPORT_SYMBOL(seq_lseek);
int seq_release(struct inode *inode, struct file *file)
{
struct seq_file *m = file->private_data;
- kfree(m->buf);
+ is_vmalloc_addr(m->buf) ? vfree(m->buf) : kfree(m->buf);
kfree(m);
return 0;
}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/