[Linux-cachefs] how to patch vanilla 2.6.26

[Michael James]

Working with http://people.redhat.com/~dhowells/nfs/nfs+fscache-39.tar.bz2
Last night I tried patchset 39 against  linux-2.6.26  and linux-2.6.26.2

linux-2.6.26.2 threw up a bit of fuzzing and offset,
 but all hunks went on apart from:
	06-cred-neuter-sys_capset.diff
	20-cred-cow-creds.diff
They are out of step with  kernel/capability.c

The patchset seems closer to plain  linux-2.6.26, 
It accepted them completely cleanly
 except for files 06 and 20 which clashed over  kernel/capability.c

== ../patchset/06-cred-neuter-sys_capset.diff ==                                       
patching file fs/open.c                                                                
patching file include/linux/security.h                                                 
patching file kernel/capability.c                                                      
Reversed (or previously applied) patch detected!  Assume -R? [n]                       
Apply anyway? [n] y                                                                    
Hunk #1 FAILED at 115.                                                                 
Hunk #2 FAILED at 149.                                                                 
Hunk #3 succeeded at 125 (offset -117 lines).                                          
Hunk #4 FAILED at 141.                                                                 
Hunk #5 succeeded at 156 with fuzz 2 (offset -113 lines).                              
Hunk #6 FAILED at 184.                                                                 
4 out of 6 hunks FAILED -- saving rejects to file kernel/capability.c.rej


>From both the patch output and a bit of eyeballing
 it seems 4 of the 6 hunks of  06  are mostly in  linux-2.6.26
But not literally enough for 20 to apply at all cleanly.

== ../patchset/20-cred-cow-creds.diff ==                                               
patching file fs/exec.c                                                                
patching file fs/nfsd/auth.c                                                           
patching file fs/nfsd/nfs4recover.c                                                    
patching file fs/open.c                                                                
patching file include/linux/capability.h                                               
patching file include/linux/cred.h                                                     
patching file include/linux/init_task.h                                                
patching file include/linux/key.h                                                      
patching file include/linux/sched.h                                                    
patching file include/linux/security.h                                                 
patching file init/main.c                                                              
patching file kernel/capability.c                                                      
Hunk #2 FAILED at 111.                                                                 
Hunk #3 FAILED at 125.                                                                 
Hunk #4 FAILED at 134.                                                                 
Hunk #5 FAILED at 167.                                                                 
Hunk #6 FAILED at 228.                                                                 
Hunk #7 succeeded at 299 (offset 56 lines).                                            
Hunk #8 FAILED at 323.                                                                 
6 out of 8 hunks FAILED -- saving rejects to file kernel/capability.c.rej              

I'm tempted to wade in and try and merge them,
 but I don't even know C, let alone what the code is doing.

I might move on and get a bit of experience
 configuring and compiling vanilla and stock kernels,
 then come back to try patched flavors.

I want to get this working on a cluster of 66 nodes
 working on bioinformatic problems using NCBI blast.
The set of files I work against no longer fit in the local disk,
 and if all the nodes start going back to the NFS server
 for the big files, the whole cluster slows to a crawl.
Fortunately a big job is always against a single fileset (~8Gig),
 so having local cache will give me the same boost
 as having a local copy.

Thanks for the work on this, I don't think it's far from flying,
michaelj

-- 
Michael James                         michael.james at csiro.au
System Administrator                    voice:  02 6246 5040
CSIRO Bioinformatics Facility             fax:  02 6246 5166

No matter how much you pay for software,
 you always get less than you hoped.
Unless you pay nothing, then you get more.