Skip to contentSkip to conversation

systemd-213: regression with zram

systemd-devel Minchan Kim 2 participants 4 messages Jun 24, 2014 – Jun 25, 2014 7 messages not in the archive

Subsystems (3)
lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): [email protected]

Minchan Kim

author maintainer
#7 lore raw
Quoted 26 lines · Alexander E. Patrakov
Quoted 4 lines · Minchan Kim

Hello,

Sorry for the late response.

Better late than never.

Quoted 6 lines

If I parse your problem correctly, you meant that changing disksize is failed while someone opens /dev/zram0?

I tried test which opens "/dev/zram0" with O_RDWR and sleep forever and then echo $((512<<20)) > /sys/block/zram0/disksize is successful. IOW, it's okay with me.

No, this is only a part of the problem. A valid test would be:

0. Reset the unused zram device.
1. Use a program that opens /dev/zram0 with O_RDWR and sleeps until killed.
2. While that program sleeps, echo the correct value to /sys/block/zram0/disksize.
3. Verify (e.g. in /proc/partitions) that the disk size is applied correctly. It is.
4. While that program still sleeps, attempt to mkswap /dev/zram0. This fails:

Quoted 3 lines

mkswap: error: swap area needs to be at least 40 KiB

There is nothing in dmesg.

Thanks. I could reproduce.
When I read strace result, fstat access closed fd which seem to be a culprit.

Cced util-linux.

<snip>

open("/dev/zram0", O_RDONLY) = 4
uname({sys="Linux", node="bboxv", ...}) = 0
ioctl(4, BLKGETSIZE64, 0x7fffa2384888) = 0
close(4) = 0
open("/usr/share/locale/locale.alias", O_RDONLY|O_CLOEXEC) = 4
fstat(4, {st_mode=S_IFREG|0644, st_size=2570, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f8a058ea000 read(4, "# Locale name alias data base.\n#"..., 4096) = 2570

read(4, "", 4096)                       = 0
close(4)                                = 0
munmap(0x7f8a058ea000, 4096)            = 0

open("/usr/share/locale/en_US/LC_MESSAGES/util-linux.mo", O_RDONLY) = -1 ENOENT (No such file or directory) open("/usr/share/locale/en/LC_MESSAGES/util-linux.mo", O_RDONLY) = -1 ENOENT (No such file or directory) open("/usr/share/locale-langpack/en_US/LC_MESSAGES/util-linux.mo", O_RDONLY) = -1 ENOENT (No such file or directory) open("/usr/share/locale-langpack/en/LC_MESSAGES/util-linux.mo", O_RDONLY) = -1 ENOENT (No such file or directory)
write(2, "mkswap: ", 8mkswap: ) = 8
write(2, "error: swap area needs to be at "..., 44error: swap area needs to be at least 40 KiB) = 44 write(2, "\n", 1
) = 1
write(2, "\nUsage:\n mkswap [options] device"..., 40
Usage:
mkswap [options] device [size]
) = 40
write(2, "\nOptions:\n -c, --check "..., 479
Options:

 -c, --check               check bad blocks before creating the swap area
 -f, --force               allow swap size area be larger than device
 -p, --pagesize SIZE       specify page size in bytes
 -L, --label LABEL         specify label
 -v, --swapversion NUM     specify swap-space version number
 -U, --uuid UUID           specify the uuid to use
 -V, --version             output version information and exit
 -h, --help                display this help and exit

) = 479
exit_group(1) = ?

But when it was successful, it was as follows via stat("/dev/zram0").

<snip>

open("/dev/zram0", O_RDONLY) = 4
uname({sys="Linux", node="bboxv", ...}) = 0
ioctl(4, BLKGETSIZE64, 0x7fff85926088) = 0
close(4) = 0
stat("/dev/zram0", {st_mode=S_IFBLK|0660, st_rdev=makedev(252, 0), ...}) = 0
open("/dev/zram0", O_RDWR|O_EXCL) = 4
open("/etc/mtab", O_RDONLY|O_CLOEXEC) = 5
fstat(5, {st_mode=S_IFREG|0644, st_size=725, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f984f537000 read(5, "/dev/vda1 / ext4 rw,errors=remou"..., 4096) = 725

read(5, "", 4096)                       = 0
close(5)                                = 0
munmap(0x7f984f537000, 4096)            = 0
ioctl(4, BLKALIGNOFF, 0x7fff85925fcc)   = 0
lseek(4, 0, SEEK_SET)                   = 0
ioctl(4, 0x301, 0x7fff85925fc0)         = -1 ENOTTY (Inappropriate ioctl for device)
fadvise64(4, 0, 0, POSIX_FADV_RANDOM)   = 0
fstat(4, {st_mode=S_IFBLK|0660, st_rdev=makedev(252, 0), ...}) = 0
uname({sys="Linux", node="bboxv", ...}) = 0
ioctl(4, BLKGETSIZE64, 0x15f72b0)       = 0
ioctl(4, CDROM_GET_CAPABILITY, 0)       = -1 ENOTTY (Inappropriate ioctl for device)
lseek(4, 0, SEEK_SET)                   = 0

read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 1024) = 1024
ioctl(4, BLKSSZGET, 0x15f72c8) = 0
lseek(4, 15872, SEEK_SET) = 15872
read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 512) = 512
lseek(4, 28672, SEEK_SET) = 28672
read(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 1024) = 1024
lseek(4, 0, SEEK_SET) = 0
write(4, "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0"..., 1024) = 1024 open("/usr/share/locale/locale.alias", O_RDONLY|O_CLOEXEC) = 5
fstat(5, {st_mode=S_IFREG|0644, st_size=2570, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f984f537000 read(5, "# Locale name alias data base.\n#"..., 4096) = 2570

read(5, "", 4096)                       = 0
close(5)                                = 0
munmap(0x7f984f537000, 4096)            = 0

open("/usr/share/locale/en_US/LC_MESSAGES/util-linux.mo", O_RDONLY) = -1 ENOENT (No such file or directory) open("/usr/share/locale/en/LC_MESSAGES/util-linux.mo", O_RDONLY) = -1 ENOENT (No such file or directory) open("/usr/share/locale-langpack/en_US/LC_MESSAGES/util-linux.mo", O_RDONLY) = -1 ENOENT (No such file or directory) open("/usr/share/locale-langpack/en/LC_MESSAGES/util-linux.mo", O_RDONLY) = -1 ENOENT (No such file or directory) fstat(1, {st_mode=S_IFREG|0644, st_size=8472, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7f984f537000
lseek(4, 1024, SEEK_SET) = 1024
write(4, "\1\0\0\0\377\377\0\0\0\0\0\0\205\236:\36b\377N\324\255\224\5\341CH(\376\0\0\0\0"..., 3072) = 3072
fsync(4) = 0
write(1, "Setting up swapspace version 1, "..., 102Setting up swapspace version 1, size = 262140 KiB no label, UUID=859e3a1e-62ff-4ed4-ad94-05e1434828fe
) = 102
exit_group(0) = ?

Quoted 3 lines
Signature · 2 lines
-- 
Alexander E. Patrakov

_______________________________________________
systemd-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/systemd-devel

author maintainer
#8 lore raw
Quoted 29 lines · Minchan Kim
Quoted 28 lines · Alexander E. Patrakov
Quoted 4 lines · Minchan Kim

Hello,

Sorry for the late response.

Better late than never.

Quoted 6 lines

If I parse your problem correctly, you meant that changing disksize is failed while someone opens /dev/zram0?

I tried test which opens "/dev/zram0" with O_RDWR and sleep forever and then echo $((512<<20)) > /sys/block/zram0/disksize is successful. IOW, it's okay with me.

No, this is only a part of the problem. A valid test would be:

0. Reset the unused zram device.
1. Use a program that opens /dev/zram0 with O_RDWR and sleeps until killed.
2. While that program sleeps, echo the correct value to /sys/block/zram0/disksize.
3. Verify (e.g. in /proc/partitions) that the disk size is applied correctly. It is.
4. While that program still sleeps, attempt to mkswap /dev/zram0. This fails:

mkswap: error: swap area needs to be at least 40 KiB

Quoted 7 lines
Quoted 1 line

There is nothing in dmesg.

Thanks. I could reproduce.
When I read strace result, fstat access closed fd which seem to be a culprit.

Cced util-linux.

When I read disk-utils/mkswap.c, it's not a problem of mkswap. The problem is ioctl(fd, BLKGETSIZE64, bytes) is not work in your scenario. IOW, kernel fault.
Hmm,, will investgate it.

Thanks for the report!
_______________________________________________
systemd-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/systemd-devel

author maintainer
#9 lore raw

Hello Alexander,

Quoted 36 lines · Minchan Kim
Quoted 35 lines · Minchan Kim
Quoted 26 lines · Alexander E. Patrakov
Quoted 4 lines · Minchan Kim

Hello,

Sorry for the late response.

Better late than never.

Quoted 6 lines

If I parse your problem correctly, you meant that changing disksize is failed while someone opens /dev/zram0?

I tried test which opens "/dev/zram0" with O_RDWR and sleep forever and then echo $((512<<20)) > /sys/block/zram0/disksize is successful. IOW, it's okay with me.

No, this is only a part of the problem. A valid test would be:

0. Reset the unused zram device.
1. Use a program that opens /dev/zram0 with O_RDWR and sleeps until killed.
2. While that program sleeps, echo the correct value to /sys/block/zram0/disksize.
3. Verify (e.g. in /proc/partitions) that the disk size is applied correctly. It is.
4. While that program still sleeps, attempt to mkswap /dev/zram0. This fails:

Quoted 3 lines

mkswap: error: swap area needs to be at least 40 KiB

There is nothing in dmesg.

Thanks. I could reproduce.
When I read strace result, fstat access closed fd which seem to be a culprit.

Quoted 8 lines
Quoted 1 line

Cced util-linux.

When I read disk-utils/mkswap.c, it's not a problem of mkswap. The problem is ioctl(fd, BLKGETSIZE64, bytes) is not work in your scenario. IOW, kernel fault.
Hmm,, will investgate it.

Thanks for the report!

Could you test this patch? It passed my test.
Thanks!

Quoted 1 line

From 4ab4931c3fa7e6527e3a849d5e4c4f727143e66c Mon Sep 17 00:00:00 2001

From: Minchan Kim <[email protected]>
Date: Wed, 25 Jun 2014 09:20:24 +0900
Subject: [PATCH] zram: revalidate disk after capacity change

Alexander reported mkswap on /dev/zram0 is failed if other process is opening the block device file.

Step is as follows,

0. Reset the unused zram device.
1. Use a program that opens /dev/zram0 with O_RDWR and sleeps until killed.
2. While that program sleeps, echo the correct value to /sys/block/zram0/disksize.
3. Verify (e.g. in /proc/partitions) that the disk size is applied correctly. It is.
4. While that program still sleeps, attempt to mkswap /dev/zram0. This fails:
result: mkswap: error: swap area needs to be at least 40 KiB

When I investigated, the size get by ioctl(fd, BLKGETSIZE64) on mkswap to get a size of blockdev was zero although zram0 has right size by 2.

The reason is zram didn't revalidate disk after changing capacity so that size of blockdev's inode is not uptodate until all of file is close.

This patch should fix the problem.

---
 drivers/block/zram/zram_drv.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)
drivers/block/zram/zram_drv.c +4 -1
index 48eccb350180..089e72cd37be 100644
@@ -622,8 +622,10 @@ static void zram_reset_device(struct zram *zram, bool reset_capacity)
 	memset(&zram->stats, 0, sizeof(zram->stats));
 
 	zram->disksize = 0;
-	if (reset_capacity)
+	if (reset_capacity) {
 		set_capacity(zram->disk, 0);
+		revalidate_disk(zram->disk);
+	}
 	up_write(&zram->init_lock);
 }
 
@@ -664,6 +666,7 @@ static ssize_t disksize_store(struct device *dev,
 	zram->comp = comp;
 	zram->disksize = disksize;
 	set_capacity(zram->disk, zram->disksize >> SECTOR_SHIFT);
+	revalidate_disk(zram->disk);
 	up_write(&zram->init_lock);
 	return len;
 
Signature · 6 lines
2.0.0

_______________________________________________
systemd-devel mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/systemd-devel
#10 lore raw
Quoted 24 lines · Minchan Kim

Could you test this patch? It passed my test.
Thanks!

From 4ab4931c3fa7e6527e3a849d5e4c4f727143e66c Mon Sep 17 00:00:00 2001
From: Minchan Kim <[email protected]>
Date: Wed, 25 Jun 2014 09:20:24 +0900
Subject: [PATCH] zram: revalidate disk after capacity change

Alexander reported mkswap on /dev/zram0 is failed if other process is opening the block device file.

Step is as follows,

0. Reset the unused zram device.
1. Use a program that opens /dev/zram0 with O_RDWR and sleeps until killed.
2. While that program sleeps, echo the correct value to /sys/block/zram0/disksize.
3. Verify (e.g. in /proc/partitions) that the disk size is applied correctly. It is.
4. While that program still sleeps, attempt to mkswap /dev/zram0. This fails:
result: mkswap: error: swap area needs to be at least 40 KiB

When I investigated, the size get by ioctl(fd, BLKGETSIZE64) on mkswap to get a size of blockdev was zero although zram0 has right size by 2.

Quoted 7 lines

The reason is zram didn't revalidate disk after changing capacity so that size of blockdev's inode is not uptodate until all of file is close.

This patch should fix the problem.

It indeed fixes the problem.

Quoted 29 lines
---
  drivers/block/zram/zram_drv.c | 5 ++++-
  1 file changed, 4 insertions(+), 1 deletion(-)
drivers/block/zram/zram_drv.c +4 -1
index 48eccb350180..089e72cd37be 100644
@@ -622,8 +622,10 @@ static void zram_reset_device(struct zram *zram, bool reset_capacity)
  	memset(&zram->stats, 0, sizeof(zram->stats));
 
  	zram->disksize = 0;
-	if (reset_capacity)
+	if (reset_capacity) {
  		set_capacity(zram->disk, 0);
+		revalidate_disk(zram->disk);
+	}
  	up_write(&zram->init_lock);
  }
 
@@ -664,6 +666,7 @@ static ssize_t disksize_store(struct device *dev,
  	zram->comp = comp;
  	zram->disksize = disksize;
  	set_capacity(zram->disk, zram->disksize >> SECTOR_SHIFT);
+	revalidate_disk(zram->disk);
  	up_write(&zram->init_lock);
  	return len;
 
Signature · 1 line
Alexander E. Patrakov