In copy_write(), if there was an error on the read portion of the copy, should we always bail out and stop the job, or just when the caller has specified a callback notifier? The current code will continue with the write portion of the copy even on an error in this situation. I'm not sure what the original intention was. --- linux-2.4.20a/drivers/md/kcopyd.c 2003/01/09 19:43:41 +++ linux-2.4.20b/drivers/md/kcopyd.c 2003/01/09 19:51:11 @@ -655,8 +655,10 @@ { struct copy_info *info = (struct copy_info *) job->context; - if (job->err && info->notify) { - info->notify(job->err, job->context); + if (job->err) { + if (info->notify) { + info->notify(job->err, job->context); + } kcopyd_free_job(job); free_copy_info(info); return;