Started by Jeff Garzik on Mar 12, 2011 · 5 messages from 4 authors · Last reply Mar 14, 2011
open the first message on its own page
J Jeff Garzik #
On 03/11/2011 10:50 PM, Indan Zupancic wrote:
If sync(2) didn't exist and people wanted to add it I'd complain too. This has all the problems of sync(2), but with the "not sure if all the files are on the file system I think" problem added.
You are decades too late, then... sync_file_range() is not appropriate because that works on only one fd. The new syscall is fine, and addresses a need. Jeff
I Indan Zupancic replied to Jeff Garzik #
On Sat, March 12, 2011 19:31, Jeff Garzik wrote:
On 03/11/2011 10:50 PM, Indan Zupancic wrote:quoted
If sync(2) didn't exist and people wanted to add it I'd complain too. This has all the problems of sync(2), but with the "not sure if all the files are on the file system I think" problem added.You are decades too late, then...
I know.
sync_file_range() is not appropriate because that works on only one fd. The new syscall is fine, and addresses a need.
The new syscall works on only one fd too. The behaviour of the proposed
syncfs and an extended sync_file_range is exactly the same.
Greetings,
IndanT Theodore Tso replied to Indan Zupancic #
On Mar 13, 2011, at 9:31 PM, Indan Zupancic wrote:
The new syscall works on only one fd too. The behaviour of the proposed syncfs and an extended sync_file_range is exactly the same.
No, it's quite different. One syncs the entire file system; sync_file_range requests that the pages associated with a file be pushed to disk --- it says nothing about the metadata associated with the file. I'm in favor of the new sys_syncfs system call. -- Ted
J Jeff Garzik replied to Indan Zupancic #
On 03/13/2011 09:31 PM, Indan Zupancic wrote:
The new syscall works on only one fd too. The behaviour of the proposed syncfs and an extended sync_file_range is exactly the same.
No, it's not. You should read the patch before commenting. Jeff
I Indan Zupancic replied to Theodore Tso #
On Mon, March 14, 2011 02:37, Theodore Tso wrote:
On Mar 13, 2011, at 9:31 PM, Indan Zupancic wrote:quoted
The new syscall works on only one fd too. The behaviour of the proposed syncfs and an extended sync_file_range is exactly the same.No, it's quite different. One syncs the entire file system; sync_file_range requests that the pages associated with a file be pushed to disk --- it says nothing about the metadata associated with the file.
You missed the part where a new flag for sync_file_range would be added that
does sync the whole file system.
I'm in favor of the new sys_syncfs system call.
Care to define its behaviour? Does it wait till all data is written out,
does it flush disk caches, what if new data is written to the fs while
a sync is going on, etc.
Greetings,
Indan