| To: | linux-xfs@xxxxxxxxxxx |
|---|---|
| Subject: | RE: guaranteed increasing extent allocation |
| From: | "Rick Smith" <rgsmith72@xxxxxxxxxxx> |
| Date: | Tue, 01 Jul 2003 11:18:02 -0700 |
| Sender: | linux-xfs-bounce@xxxxxxxxxxx |
|
Hello, I have made some progress concerning this issue with slight modifications to the XFS source code. Specifically, changing the default allocation type to XFS_ALLOCTYPE_FIRST_AG and set the default allocation group to zero (when nullfb == NULL in xfs_bmap_alloc() in xfs_bmap.c). It is my hope to always force allocation to start searching at the beginning of allocation group 0 and then proceed to the next group once the preceding group is full. Since I am using XFS_IOC_RESVSP64 to reserve the files ahead of time, I am not concerned with potentially slowing down the file creation process. I also realize that this is probably not what most filesystem users are looking for, but I think it is crucial for video I/O. Is there a way to specify to the allocator to always begin with the start of an allocation group? There are cases where a file gets allocated from an allocation group and then a later file allocation gets a space closer to the beginning of the same allocation group. Is this due to the binary block search in xfs_alloc_ag_vextent_size()? Also, are there any other potential drawbacks to setting the default allocation type to XFS_ALLOCTYPE_FIRST_AG? Thanks. Rick Smith _________________________________________________________________ MSN 8 helps eliminate e-mail viruses. Get 2 months FREE*. http://join.msn.com/?page=features/virus |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | TAKE - [2.5] Remove dead xfs_syncd.c file, Eric Sandeen |
|---|---|
| Next by Date: | oops in 2.4.20-18.9XFS1.3.0pre2, Sebastien Boving |
| Previous by Thread: | TAKE - [2.5] Remove dead xfs_syncd.c file, Eric Sandeen |
| Next by Thread: | oops in 2.4.20-18.9XFS1.3.0pre2, Sebastien Boving |
| Indexes: | [Date] [Thread] [Top] [All Lists] |