Bug 46361

Summary: ext4 partition mount takes long time with 3.5.1+ kernel
Product: File System Reporter: nzqr (candysnell)
Component: ext4Assignee: fs_ext4 (fs_ext4)
Status: RESOLVED CODE_FIX    
Severity: normal CC: alan, tytso
Priority: P1    
Hardware: All   
OS: Linux   
Kernel Version: 3.5.1 Subsystem:
Regression: Yes Bisected commit-id:
Attachments: my kernel config
dumpe2fs -h report of this partition
dmesg of kernel on which bug happens
smartctl -a for hdd on which this happens - no errors
hdparm -I of this hdd - all seems correct

Description nzqr 2012-08-23 12:05:14 UTC
Created attachment 78261 [details]
my kernel config

First to say, this happens with 3.5.2 kernel too. After upgrade to 3.5.1 mount of my 200G+ partition started taking a lot of time, here's `strace -tt mount -a` with 3.5.1 kernel:

15:45:26.074958 mount("/dev/sda3", "/mnt/2", "ext4", MS_MGC_VAL|MS_NOATIME, "errors=remount-ro,data=journal") = 0
15:45:34.940793 access("/mnt/2", W_OK)  = 0

and with 3.5.0 with same config:

15:09:33.408747 mount("/dev/sda3", "/mnt/2", "ext4", MS_MGC_VAL|MS_NOATIME, "errors=remount-ro,data=journal") = 0
15:09:33.578511 access("/mnt/2", W_OK)  = 0

First I thought it was corruption on fs, but fsck -f -v and even formatting doesn't fix that. And, seems that small partition (4G), where my / is, still mounts fast.
Comment 1 nzqr 2012-08-23 12:06:54 UTC
Created attachment 78271 [details]
dumpe2fs -h report of this partition
Comment 2 nzqr 2012-08-23 12:07:57 UTC
Created attachment 78281 [details]
dmesg of kernel on which bug happens
Comment 3 nzqr 2012-08-23 12:13:51 UTC
Created attachment 78291 [details]
smartctl -a for hdd on which this happens - no errors
Comment 4 nzqr 2012-08-23 12:14:35 UTC
Created attachment 78301 [details]
hdparm -I of this hdd - all seems correct
Comment 5 nzqr 2012-08-23 12:37:02 UTC
And yes, this happens even without ugly tainting nvidia's module.
Comment 6 Theodore Tso 2012-08-23 19:16:43 UTC
This was fixed in commit 0548bbb85337e532ca2ed697c3e9b227ff2ed4b4 which hit mainline in v3.6-rc3.  It will hopefully be in the next stable release, v3.5.3, but that hasn't been released yet.

Please try cherry-picking that fix, and let me know if that doesn't fix it.  If it doesn't, we can reopen this bug and I'll take a closer look.  But I'm 99.9% sure this should fix things for you.