Hi Christoph, On Sat, Jun 25, 2022 at 06:10:02AM -0700, Christoph Hellwig wrote: > On Sat, Jun 25, 2022 at 01:01:13PM +0200, Jason A. Donenfeld wrote: > > Now that all callers of ->llseek are going through vfs_llseek(), we > > don't gain anything by keeping no_llseek around. Nothing compares it or > > calls it. > > Shouldn't this and the checks for no_llseek simply be merged into patch > 2? I'd done that at first, but Al had suggested it be a separate commit in <https://lore.kernel.org/lkml/YrYxOC5dgCKBHwVE@ZenIV/>, when he mentions "next commit would", so I did how he asked. > > > + if ((file->f_mode & FMODE_LSEEK) && file->f_op->llseek) > > + return file->f_op->llseek(file, offset, whence); > > + return -ESPIPE; > > No function change, but in general checking for the error condition > in the branch tends to be more readable. i.e.: > > if (!(file->f_mode & FMODE_LSEEK) || !file->f_op->llseek) > return -ESPIPE; > return file->f_op->llseek(file, offset, whence); > I thought about this kind of reverse: what is the acceptable condition in which one may call ->llseek? Easier to express it that way than in the inverse. But if you really want, I can change it around if there's a v3 with other changes (which at the moment doesn't seem like there's going to be). Jason
