FIO bug

FIO bug

[H. Tolley]

The function static void get_file_system_type(struct fio_file *f)�� (in file filesetup.c)
has a call to lstat(..) which if given a symbolic link will give information about the link 
instead of information about the target of the link.� In Solaris it is very common that the 
disk device nodes in /dev/dsk/ are symbolic links to files in /devices/... and so the call 
to lstat(..) will not properly identify a file as a block device instead of a normal file.� 

A call to stat(..) on a filename that represents a block device will not return the proper 
size on Solaris.� Combined with the above-mentioned mis-identification of a link to a 
block device as a normal file, the function file_size(..) is invoked rather than a call
to bdev_size(..) and fio will not run properly.

Fix: Can the call to lstat(..) in get_file_system_type(..) be replaced by a call to stat(..)?

This happens on Open Solaris v10 x86_64.

Hal Tolley
 		 	   		  
_________________________________________________________________
Your E-mail and More On-the-Go. Get Windows Live Hotmail Free.
http://clk.atdmt.com/GBL/go/201469229/direct/01/

Re: FIO bug

[Jens Axboe]

On Fri, Feb 05 2010, H. Tolley wrote:
> 
> The function static void get_file_system_type(struct fio_file *f)��
> (in file filesetup.c) has a call to lstat(..) which if given a
> symbolic link will give information about the link instead of
> information about the target of the link.� In Solaris it is very
> common that the disk device nodes in /dev/dsk/ are symbolic links to
> files in /devices/... and so the call to lstat(..) will not properly
> identify a file as a block device instead of a normal file.� 
> 
> A call to stat(..) on a filename that represents a block device will
> not return the proper size on Solaris.� Combined with the
> above-mentioned mis-identification of a link to a block device as a
> normal file, the function file_size(..) is invoked rather than a call
> to bdev_size(..) and fio will not run properly.
> 
> Fix: Can the call to lstat(..) in get_file_system_type(..) be replaced
> by a call to stat(..)?
> 
> This happens on Open Solaris v10 x86_64.

That does look like a bug, that stat should return info about the real
device and not a potential link. I have committed a fix for that now.

-- 
Jens Axboe

Re: fio bug

[Sitsofe Wheeler]

On 27 June 2016 at 22:20, Frank Londono <flonclondono@gmail.com> wrote:
> Just started using tkperf/fio, and am the system crashes:
>
> Jun 27 12:02:14 qual6 org.debian.apt[1424]: 12:02:14 AptDaemon.Worker
> [INFO]: Finished transaction /org/debian/apt/transaction/
> Jun 27 12:02:44 qual6 kernel: [ 1939.769050] swap_free: Bad swap
> offset entry 8641b9fff80
> Jun 27 12:02:44 qual6 kernel: [ 1939.769059] BUG: Bad page map in
> process fio  pte:10c8373fff0040 pmd:335c5067
> Jun 27 12:02:44 qual6 kernel: [ 1939.769063] addr:00007ff321a00000
> vm_flags:080000fb anon_vma:          (null) mapping:ffff880196cc8e90
> index:3ab2
> Jun 27 12:02:44 qual6 kernel: [ 1939.769073] file:SYSV00000000
> fault:shm_fault mmap:shm_mmap readpage:          (null)
> Jun 27 12:02:44 qual6 kernel: [ 1939.769078] CPU: 0 PID: 3005 Comm:
> fio Not tainted 4.4.0-24-generic #43-Ubuntu
> Jun 27 12:02:44 qual6 kernel: [ 1939.769081] Hardware name: Dell
> Computer Corporation PowerEdge 1800/0P8611, BIOS A07 09/29/2006
> Jun 27 12:02:44 qual6 kernel: [ 1939.769084]  0000000000000286
> 0000000063f6c1a4 ffff880196f97c38 ffffffff813eab23
> Jun 27 12:02:44 qual6 kernel: [ 1939.769090]  00007ff321a00000
> ffff880196eb2578 ffff880196f97c88 ffffffff811bc32f
> Jun 27 12:02:44 qual6 kernel: [ 1939.769095]  ffff880196f97c60
> ffffffff811d33e0 000008641b9fff80 0010c8373fff0040
> Jun 27 12:02:44 qual6 kernel: [ 1939.769099] Call Trace:
> Jun 27 12:02:44 qual6 kernel: [ 1939.769109]  [<ffffffff813eab23>]
> dump_stack+0x63/0x90
> Jun 27 12:02:44 qual6 kernel: [ 1939.769115]  [<ffffffff811bc32f>]
> print_bad_pte+0x1df/0x2a0
> Jun 27 12:02:44 qual6 kernel: [ 1939.769121]  [<ffffffff811d33e0>] ?
> swap_info_get+0xb0/0xd0
> Jun 27 12:02:44 qual6 kernel: [ 1939.769125]  [<ffffffff811bdcb8>]
> unmap_page_range+0x468/0x7a0
> Jun 27 12:02:44 qual6 kernel: [ 1939.769129]  [<ffffffff811be06d>]
> unmap_single_vma+0x7d/0xe0
> Jun 27 12:02:44 qual6 kernel: [ 1939.769133]  [<ffffffff811beb31>]
> unmap_vmas+0x51/0xa0
> Jun 27 12:02:44 qual6 kernel: [ 1939.769136]  [<ffffffff811c45bc>]
> unmap_region+0xac/0x120
> Jun 27 12:02:44 qual6 kernel: [ 1939.769141]  [<ffffffff81223cd0>] ?
> __dentry_kill+0x160/0x1e0
> Jun 27 12:02:44 qual6 kernel: [ 1939.769145]  [<ffffffff811c6aca>]
> do_munmap+0x28a/0x480
> Jun 27 12:02:44 qual6 kernel: [ 1939.769149]  [<ffffffff8132c9d2>]
> SyS_shmdt+0xb2/0x170
> Jun 27 12:02:44 qual6 kernel: [ 1939.769156]  [<ffffffff81825bf2>]
> entry_SYSCALL_64_fastpath+0x16/0x71
> Jun 27 12:02:44 qual6 kernel: [ 1939.769159] Disabling lock debugging
> due to kernel taint
> Jun 27 12:02:44 qual6 kernel: [ 1939.769161] swap_free: Bad swap
> offset entry 1f800000

Assuming your hardware isn't buggy you've tickled a bug in the kernel
rather than a bug in fio so this isn't the right list of your issue.
Since you're running an Ubuntu kernel you need to report it to Ubuntu
(e.g. via https://bugs.launchpad.net/ubuntu/+source/linux/+filebug )
unless you can also reproduce it on the latest mainline kernel from
kernel.org (4.7-rc5 at the time of writing) in which case you would
better off reporting on the Linux Kernel Mailing List / the kernel.org
Bugzilla.

-- 
Sitsofe | http://sucs.org/~sits/