Re: [PATCH] fs: kill default_llseek

[Josef Bacik <josef@redhat.com>]

On 05/09/2011 02:30 PM, Dave Anderson wrote:
>
>
> ----- Original Message -----
>> On 05/09/2011 02:03 PM, Dave Anderson wrote:
>>>
>>> On Thursday 05 May 2011 16:27:57 Josef Bacik wrote:
>>>> Looking at this llseek stuff I noticed that default_llseek is the
>>>> exact same as
>>>> generic_file_llseek, so kill default_llseek. I patched this using
>>>> spatch with
>>>> just a simple
>>>>
>>>> @@
>>>> @@
>>>>
>>>> - default_llseek
>>>> + generic_file_llseek
>>>
>>> ...
>>>
>>>> diff --git a/fs/proc/kcore.c b/fs/proc/kcore.c
>>>> index d245cb2..6f37c39 100644
>>>> --- a/fs/proc/kcore.c
>>>> +++ b/fs/proc/kcore.c
>>>> @@ -558,7 +558,7 @@ static int open_kcore(struct inode *inode,
>>>> struct file *filp)
>>>>    static const struct file_operations proc_kcore_operations = {
>>>>    	.read = read_kcore,
>>>>    	.open = open_kcore,
>>>> - .llseek = default_llseek,
>>>> + .llseek = generic_file_llseek,
>>>>    };
>>>
>>>> diff --git a/fs/proc/vmcore.c b/fs/proc/vmcore.c
>>>> index 74802bc5..0cafd9e 100644
>>>> --- a/fs/proc/vmcore.c
>>>> +++ b/fs/proc/vmcore.c
>>>> @@ -163,7 +163,7 @@ static ssize_t read_vmcore(struct file *file,
>>>> char __user *buffer,
>>>>
>>>>    static const struct file_operations proc_vmcore_operations = {
>>>>    	.read = read_vmcore,
>>>> - .llseek = default_llseek,
>>>> + .llseek = generic_file_llseek,
>>>>    };
>>>
>>> Both /proc/kcore and /proc/vmcore currently require
>>> default_llseek().
>>> They were both changed to use generic_file_llseek(), but then
>>> subsequently
>>> reverted back to default_llseek():
>>>
>>> commit c227e69028473c7c7994a9b0a2cc0034f3f7e0fe
>>> Author: Arnd Bergmann<arnd@arndb.de>
>>> Date: Wed Sep 22 13:04:54 2010 -0700
>>>
>>>       /proc/vmcore: fix seeking
>>>
>>>       Commit 73296bc611 ("procfs: Use generic_file_llseek in
>>>       /proc/vmcore")
>>>       broke seeking on /proc/vmcore. This changes it back to use
>>>       default_llseek
>>>       in order to restore the original behaviour.
>>>       ...
>>>
>>>
>>> commit ceff1a770933e2ca2bf995b453dade4ec47a9878
>>> Author: Dave Anderson<anderson@redhat.com>
>>> Date: Wed Jan 12 17:00:36 2011 -0800
>>>
>>>       /proc/kcore: fix seeking
>>>
>>>       Commit 34aacb2920 ("procfs: Use generic_file_llseek in
>>>       /proc/kcore") broke
>>>       seeking on /proc/kcore. This changes it back to use
>>>       default_llseek in
>>>       order to restore the original behavior.
>>>       ...
>>>
>>
>> How is it getting s_maxbytes set to 0? I'm looking everywhere and I
>> can't see how that can happen. It seems that anybody using sget should
>> be getting it set to MAX_NONLFS so they should all be ok. I'm looking
>> at proc in particular and it doesn't do anything special, so it should
>> be ok. (Obviously it wasn't, I'm just trying to understand how we're
>> getting s_maxbytes == 0 so we can fix that and kill default_llseek).
>> Thanks,
>>
>> Josef
>
> Well, in the case of /proc/kcore, it was set to MAX_NON_LFS (2GB-1)
> which is too small because the file offset values in the /proc/kcore
> PT_LOAD segments may exceed or start beyond that offset value.
>
> I guessing the same thing was seen with /proc/vmcore, even
> though Arnd's commit message implies that it was 0?
>

Ah yeah I guess that's it.  Alright I'll think of something else.  Thanks,

Josef