Re: msync(2) bug(?), returns AOP_WRITEPAGE_ACTIVATE to userland

[Erez Zadok]

That's a nice historical review, Huge, of how got into these mess we're in
now -- it all starts with good intentions. :-)

On a related note, I would just love to get rid of calling the lower
->writepage in unionfs b/c I can't even tell if I have a lower page to use
all the time.  I'd prefer to call vfs_write() if I can, but I'll need a
struct file, or at least a dentry.

What ecryptfs does is store a struct file inside it's inode, so it can use
it later in ->writepage to call vfs_write on the lower f/s.  And Unionfs may
have to go in that direction too, but this trick is not terribly clean --
storing a file inside an inode.

I realize that the calling path to ->writepage doesn't have a file/dentry
any more, but if we're considering larger changes to the writepage related
code, can we perhaps consider passing a file or dentry to >writepage (same
as commit_write, perhaps).

Thanks,
Erez.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/