In message <[email protected]>,Simon Wilkinson 
writes:
>Yes, the patch removes put_inode completely. But it also makes  
>denty_iput call afs_InactiveVCache unconditionally (previously, it  
>only called it when CUnlinked was set).

afs_InactiveVCache() already has a conditional check for CUnlinked so
i didnt think it was helpful to check twice.

>My question is whether doing so just because the kernel has reduced  
>the reference count on the dentry is correct?

.dentry_iput() is only called when the refcount on the kernel inode
drops to 1.  atleast it did the last time i checked.  only .put_inode
was called for every decrement in the kernel inode refcount.

>> i dont think afs_InactiveVCache() specifically clears any callbacks.
>> however, it probably does make the vnode ready for recycling.
>
>afs_InactiveVCache calls afs_InvalidateAllSegments if CDirty is set,  
>afs_InvalidateAllSegments clears the callback, as well as deleting any  
>data that hasn't yet been sent to the fileserver.

if .dentry_iput is called, the inode is going inactive.  however, it
might be a bit soon to start throwing away data.  are you loosing
data?
_______________________________________________
OpenAFS-devel mailing list
[email protected]
https://lists.openafs.org/mailman/listinfo/openafs-devel

Reply via email to