Hi Christoph,
I really like your refactoring. It looks a lot better than the
original lists-every-where code. One minor comment below:
<snip..>
> static void bl_write_cleanup(struct work_struct *work)
> {
> - struct rpc_task *task;
> - struct nfs_pgio_header *hdr;
> + struct rpc_task *task = container_of(work, struct rpc_task, u.tk_work);
> + struct nfs_pgio_header *hdr =
> + container_of(task, struct nfs_pgio_header, task);
> +
> dprintk("%s enter\n", __func__);
> - task = container_of(work, struct rpc_task, u.tk_work);
> - hdr = container_of(task, struct nfs_pgio_header, task);
> +
> if (likely(!hdr->pnfs_error)) {
> - /* Marks for LAYOUTCOMMIT */
> - mark_extents_written(BLK_LSEG2EXT(hdr->lseg),
> - hdr->args.offset, hdr->args.count);
> + struct pnfs_block_layout *bl = BLK_LSEG2EXT(hdr->lseg);
> + u64 start = hdr->args.offset & (loff_t)PAGE_CACHE_MASK;
> + u64 end = (hdr->args.offset + hdr->args.count +
> + PAGE_CACHE_SIZE - 1) & (loff_t)PAGE_CACHE_MASK;
> +
> + ext_tree_mark_written(bl, start >> SECTOR_SHIFT,
> + (end - start) >> SECTOR_SHIFT);
The old code works in a way that mark_extents_written() always
succeeds. Now that ext_tree_mark_written() may fail, it should check
for the return value and resend to MDS if failed, otherwise there may
be silent data corruption.
Thanks,
Tao
--
To unsubscribe from this list: send the line "unsubscribe linux-nfs" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
