From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Tue, 24 Jul 2001 12:25:11 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Tue, 24 Jul 2001 12:25:02 -0400 Received: from c1399491-a.hlndpk1.il.home.com ([24.14.108.88]:50425 "EHLO stompy.outflux.net") by vger.kernel.org with ESMTP id ; Tue, 24 Jul 2001 12:24:45 -0400 Date: Tue, 24 Jul 2001 11:24:51 -0500 From: Kees Cook To: linux-kernel@vger.kernel.org Subject: [PATCH] add "autorun" interface to md Message-ID: <20010724112451.B6287@cpoint.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Original-Recipient: rfc822;linux-kernel-outgoing I sent this to the linux-raid list, but the list appears dead? I guess I'll go here instead. Flame me nicely if I should send it elsewhere. :) After boot-up (or as a module) the "md" driver has no interface to run the "autostart_arrays" function. In the case of removable disks (eg USB, or in my case, FireWire), since the disks may not appear in the same place, or in the same order, the standard raidtools' "raidstart" will not work (calling the md.c "raidstart" interface) because the device names don't match up. While I have this sneaking worry that md's persistant superblock may be the part of this design that needs adjusting, the existing code lacks an interface to the "autorun" feature available during boot time. The interface I've made depends on the user to hand md.c the partitions to use for the auto-starting (normally this is done by the "partitions" code). I'm not sure if this is done properly (we used the proc interface example from /proc/scsi/scsi), but I'd like to at least start a discussion so I can autorun my removable raid the "right" way. ;) Thanks go to Aaron Botsis who got me started by writing the proc-write skeleton while I was agonizing over how to parse the input. For reference, this would auto-run a raid on the sdd1, sde1, sdf1, sdg1 partitions: echo "autorun 8:49 8:65 8:81 8:97" > /proc/mdstat ---begin--- --- linux-2.4.7/drivers/md/md.orig.c Wed Jul 18 09:32:13 2001 +++ linux-2.4.7/drivers/md/md.c Sat Jul 21 09:19:35 2001 @@ -52,6 +52,11 @@ #define MAJOR_NR MD_MAJOR #define MD_DRIVER +#define PROC_BLOCK_SIZE (3*1024) /* 4K page size, but our output routines + * use some slack for overruns + */ + + #include #define DEBUG 0 @@ -61,9 +66,7 @@ # define dprintk(x...) do { } while(0) #endif -#ifndef MODULE static void autostart_arrays (void); -#endif static mdk_personality_t *pers[MAX_PERSONALITY]; @@ -1086,8 +1089,10 @@ mdk_rdev_t *rdev; unsigned int size; - if (find_rdev_all(newdev)) + if (find_rdev_all(newdev)) { + printk("find_rdev_all failed for %s!\n", partition_name(newdev)); return -EEXIST; + } rdev = (mdk_rdev_t *) kmalloc(sizeof(*rdev), GFP_KERNEL); if (!rdev) { @@ -1103,8 +1108,10 @@ goto abort_free; } - if ((err = alloc_disk_sb(rdev))) + if ((err = alloc_disk_sb(rdev))) { + printk("alloc_disk_sb failed for %s!\n", partition_name(newdev)); goto abort_free; + } rdev->dev = newdev; if (lock_rdev(rdev)) { @@ -1721,15 +1728,15 @@ #define STILL_MOUNTED KERN_WARNING \ "md: md%d still mounted.\n" #define STILL_IN_USE \ -"md: md%d still in use.\n" +"md: md%d still in use(%d).\n" static int do_md_stop (mddev_t * mddev, int ro) { - int err = 0, resync_interrupted = 0; + int active, err = 0, resync_interrupted = 0; kdev_t dev = mddev_to_kdev(mddev); - if (atomic_read(&mddev->active)>1) { - printk(STILL_IN_USE, mdidx(mddev)); + if ((active=atomic_read(&mddev->active))>1) { + printk(STILL_IN_USE, mdidx(mddev), active); OUT(-EBUSY); } @@ -2777,6 +2784,7 @@ static int md_release (struct inode *inode, struct file * file) { mddev_t *mddev = kdev_to_mddev(inode->i_rdev); + if (mddev) atomic_dec(&mddev->active); return 0; @@ -3038,6 +3046,53 @@ return sz; } +static int md_status_write_proc(struct file *file, const char *buf, + unsigned long count, void *data) +{ + char *page; + char *ptr; + + if (count > PROC_BLOCK_SIZE) + return -EOVERFLOW; + + if (!(page=(char *) __get_free_page(GFP_KERNEL))) + return -ENOMEM; + + copy_from_user(page, buf, count); + + + if (count > 0 && page[count-1]=='\n') { + page[count-1]='\0'; + } + else return -EINVAL; + + if (count<11) + return -EINVAL; + + if (!strncmp("autorun ",page,8)) { + char * devstr; + char * major, * minor; + int dev; + + ptr=page+8; + + while ((devstr=strsep(&ptr," \n"))) { + printk("md: adding %s...\n",devstr); + major=strsep(&devstr,":"); + minor=strsep(&devstr,":"); + + dev=MKDEV(simple_strtoul(major,NULL,0), + simple_strtoul(minor,NULL,0)); + md_autodetect_dev(dev); + } + + autostart_arrays(); + } + + + free_page((ulong) page); + return count; +} static int md_status_read_proc(char *page, char **start, off_t off, int count, int *eof, void *data) { @@ -3489,6 +3544,7 @@ static void md_geninit (void) { int i; + struct proc_dir_entry *p; for(i = 0; i < MAX_MD_DEVS; i++) { md_blocksizes[i] = 1024; @@ -3504,7 +3560,8 @@ dprintk("md: sizeof(mdp_super_t) = %d\n", (int)sizeof(mdp_super_t)); #ifdef CONFIG_PROC_FS - create_proc_read_entry("mdstat", 0, NULL, md_status_read_proc, NULL); + p=create_proc_read_entry("mdstat", 0, NULL, md_status_read_proc, NULL); + p->write_proc=md_status_write_proc; #endif } @@ -3553,8 +3610,6 @@ } -#ifndef MODULE - /* * When md (and any require personalities) are compiled into the kernel * (not a module), arrays can be assembles are boot time using with AUTODETECT @@ -3617,6 +3672,7 @@ autorun_devices(-1); } +#ifndef MODULE static struct { char device_set [MAX_MD_DEVS]; int pers[MAX_MD_DEVS]; ---end--- -- Kees Cook @outflux.net