kernel-ark/fs
Mauricio Lin e070ad49f3 [PATCH] add /proc/pid/smaps
Add a "smaps" entry to /proc/pid: show howmuch memory is resident in each
mapping.

People that want to perform a memory consumption analysing can use it
mainly if someone needs to figure out which libraries can be reduced for
embedded systems.  So the new features are the physical size of shared and
clean [or dirty]; private and clean [or dirty].

Take a look the example below:

# cat /proc/4576/smaps

08048000-080dc000 r-xp /bin/bash
Size:               592 KB
Rss:                500 KB
Shared_Clean:       500 KB
Shared_Dirty:         0 KB
Private_Clean:        0 KB
Private_Dirty:        0 KB
080dc000-080e2000 rw-p /bin/bash
Size:                24 KB
Rss:                 24 KB
Shared_Clean:         0 KB
Shared_Dirty:         0 KB
Private_Clean:        0 KB
Private_Dirty:       24 KB
080e2000-08116000 rw-p
Size:               208 KB
Rss:                208 KB
Shared_Clean:         0 KB
Shared_Dirty:         0 KB
Private_Clean:        0 KB
Private_Dirty:      208 KB
b7e2b000-b7e34000 r-xp /lib/tls/libnss_files-2.3.2.so
Size:                36 KB
Rss:                 12 KB
Shared_Clean:        12 KB
Shared_Dirty:         0 KB
Private_Clean:        0 KB
Private_Dirty:        0 KB
...

(Includes a cleanup from "Richard Purdie" <rpurdie@rpsys.net>)

From: Torsten Foertsch <torsten.foertsch@gmx.net>

show_smap calls first show_map and then prints its additional information to
the seq_file.  show_map checks if all it has to print fits into the buffer and
if yes marks the current vma as written.  While that is correct for show_map
it is not for show_smap.  Here the vma should be marked as written only after
the additional information is also written.

The attached patch cures the problem.  It moves the functionality of the
show_map function to a new function show_map_internal that is called with an
additional struct mem_size_stats* argument.  Then show_map calls
show_map_internal with NULL as struct mem_size_stats* whereas show_smap calls
it with a real pointer.  Now the final

	if (m->count < m->size)  /* vma is copied successfully */
		m->version = (vma != get_gate_vma(task))? vma->vm_start: 0;

is done only if the whole entry fits into the buffer.

Signed-off-by: Hugh Dickins <hugh@veritas.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-09-05 00:05:49 -07:00
..
adfs
affs
afs
autofs
autofs4
befs
bfs
cifs
coda
cramfs
debugfs
devfs
devpts
efs
exportfs
ext2
ext3
fat
freevxfs
hfs
hfsplus
hostfs
hpfs
hppfs
hugetlbfs
isofs
jbd
jffs
jffs2
jfs Merge refs/heads/for-linus from master.kernel.org:/pub/scm/linux/kernel/git/shaggy/jfs-2.6.git 2005-08-30 07:47:42 -07:00
lockd
minix
msdos
ncpfs
nfs
nfs_common
nfsd
nls
ntfs
openpromfs
partitions
proc [PATCH] add /proc/pid/smaps 2005-09-05 00:05:49 -07:00
qnx4
ramfs
reiserfs
romfs
smbfs
sysfs
sysv
udf
ufs
umsdos
vfat
xfs
aio.c
attr.c
bad_inode.c
binfmt_aout.c
binfmt_elf_fdpic.c
binfmt_elf.c
binfmt_em86.c
binfmt_flat.c [PATCH] uclinux: use MAP_PRIVATE when mmaping code regions in flat binary loader 2005-09-02 00:57:31 -07:00
binfmt_misc.c
binfmt_script.c
binfmt_som.c
bio.c
block_dev.c
buffer.c
char_dev.c
compat_ioctl.c
compat.c
dcache.c
dcookies.c
direct-io.c
dnotify.c
dquot.c
eventpoll.c
exec.c
fcntl.c
fifo.c
file_table.c
file.c
filesystems.c
fs-writeback.c
inode.c
inotify.c
ioctl.c
ioprio.c
Kconfig
Kconfig.binfmt
libfs.c
locks.c
Makefile
mbcache.c
mpage.c
namei.c
namespace.c
nfsctl.c
open.c
pipe.c
posix_acl.c
quota_v1.c
quota_v2.c
quota.c
read_write.c
readdir.c
select.c
seq_file.c
stat.c
super.c
xattr_acl.c
xattr.c