--- /dev/null
+diff -urp linux-2.6.18.1.orig/block/ll_rw_blk.c linux-2.6.18.1/block/ll_rw_blk.c
+--- linux-2.6.18.1.orig/block/ll_rw_blk.c 2006-10-14 06:34:03.000000000 +0300
++++ linux-2.6.18.1/block/ll_rw_blk.c 2007-05-29 14:50:46.000000000 +0300
+@@ -2993,6 +2993,8 @@ static void handle_bad_sector(struct bio
+ set_bit(BIO_EOF, &bio->bi_flags);
+ }
+
++int dev_check_rdonly(struct block_device *bdev);
++
+ /**
+ * generic_make_request: hand a buffer to its device driver for I/O
+ * @bio: The bio describing the location in memory and on the device.
+@@ -3076,6 +3078,12 @@ end_io:
+
+ if (unlikely(test_bit(QUEUE_FLAG_DEAD, &q->queue_flags)))
+ goto end_io;
++ /* this is cfs's dev_rdonly check */
++ if (bio->bi_rw == WRITE &&
++ dev_check_rdonly(bio->bi_bdev)) {
++ bio_endio(bio, bio->bi_size, 0);
++ break;
++ }
+
+ /*
+ * If this device has partitions, remap block n
+@@ -3675,6 +3683,91 @@ void swap_io_context(struct io_context *
+ *ioc2 = temp;
+ }
+ EXPORT_SYMBOL(swap_io_context);
++ /*
++ * Debug code for turning block devices "read-only" (will discard writes
++ * silently). This is for filesystem crash/recovery testing.
++ */
++struct deventry {
++ dev_t dev;
++ struct deventry *next;
++};
++
++static struct deventry *devlist = NULL;
++static spinlock_t devlock = SPIN_LOCK_UNLOCKED;
++
++int dev_check_rdonly(struct block_device *bdev)
++{
++ struct deventry *cur;
++ if (!bdev) return 0;
++ spin_lock(&devlock);
++ cur = devlist;
++ while(cur) {
++ if (bdev->bd_dev == cur->dev) {
++ spin_unlock(&devlock);
++ return 1;
++ }
++ cur = cur->next;
++ }
++ spin_unlock(&devlock);
++ return 0;
++}
++
++void dev_set_rdonly(struct block_device *bdev)
++{
++ struct deventry *newdev, *cur;
++
++ if (!bdev)
++ return;
++ newdev = kmalloc(sizeof(struct deventry), GFP_KERNEL);
++ if (!newdev)
++ return;
++
++ spin_lock(&devlock);
++ cur = devlist;
++ while(cur) {
++ if (bdev->bd_dev == cur->dev) {
++ spin_unlock(&devlock);
++ kfree(newdev);
++ return;
++ }
++ cur = cur->next;
++ }
++ newdev->dev = bdev->bd_dev;
++ newdev->next = devlist;
++ devlist = newdev;
++ spin_unlock(&devlock);
++ printk(KERN_WARNING "Turning device %s (%#x) read-only\n",
++ bdev->bd_disk ? bdev->bd_disk->disk_name : "", bdev->bd_dev);
++}
++
++void dev_clear_rdonly(struct block_device *bdev)
++{
++ struct deventry *cur, *last = NULL;
++ if (!bdev) return;
++ spin_lock(&devlock);
++ cur = devlist;
++ while(cur) {
++ if (bdev->bd_dev == cur->dev) {
++ if (last)
++ last->next = cur->next;
++ else
++ devlist = cur->next;
++ spin_unlock(&devlock);
++ kfree(cur);
++ printk(KERN_WARNING "Removing read-only on %s (%#x)\n",
++ bdev->bd_disk ? bdev->bd_disk->disk_name :
++ "unknown block", bdev->bd_dev);
++ return;
++ }
++ last = cur;
++ cur = cur->next;
++ }
++ spin_unlock(&devlock);
++}
++
++EXPORT_SYMBOL(dev_set_rdonly);
++EXPORT_SYMBOL(dev_clear_rdonly);
++EXPORT_SYMBOL(dev_check_rdonly);
+
+ /*
+ * sysfs parts below
+diff -urp linux-2.6.18.1.orig/fs/block_dev.c linux-2.6.18.1/fs/block_dev.c
+--- linux-2.6.18.1.orig/fs/block_dev.c 2006-10-14 06:34:03.000000000 +0300
++++ linux-2.6.18.1/fs/block_dev.c 2007-05-29 14:53:38.000000000 +0300
+@@ -58,6 +58,7 @@ static void kill_bdev(struct block_devic
+ {
+ invalidate_bdev(bdev, 1);
+ truncate_inode_pages(bdev->bd_inode->i_mapping, 0);
++ dev_clear_rdonly(bdev);
+ }
+
+ int set_blocksize(struct block_device *bdev, int size)
+diff -urp linux-2.6.18.1.orig/include/linux/fs.h linux-2.6.18.1/include/linux/fs.h
+--- linux-2.6.18.1.orig/include/linux/fs.h 2006-10-14 06:34:03.000000000 +0300
++++ linux-2.6.18.1/include/linux/fs.h 2007-05-29 14:50:46.000000000 +0300
+@@ -1632,6 +1632,10 @@ extern void file_kill(struct file *f);
+ extern void submit_bio(int, struct bio *);
+ extern int bdev_read_only(struct block_device *);
+ #endif
++#define HAVE_CLEAR_RDONLY_ON_PUT
++extern void dev_set_rdonly(struct block_device *bdev);
++extern int dev_check_rdonly(struct block_device *bdev);
++extern void dev_clear_rdonly(struct block_device *bdev);
+ extern int set_blocksize(struct block_device *, int);
+ extern int sb_set_blocksize(struct super_block *, int);
+ extern int sb_min_blocksize(struct super_block *, int);
--- /dev/null
+Index: linux-2.6.22.5/drivers/scsi/Kconfig
+===================================================================
+--- linux-2.6.22.5.orig/drivers/scsi/Kconfig
++++ linux-2.6.22.5/drivers/scsi/Kconfig
+@@ -76,6 +76,14 @@ config BLK_DEV_SD
+ In this case, do not compile the driver for your SCSI host adapter
+ (below) as a module either.
+
++config SD_IOSTATS
++ bool "Enable SCSI disk I/O stats"
++ depends on BLK_DEV_SD
++ default y
++ ---help---
++ This enables SCSI disk I/O stats collection. You must also enable
++ /proc file system support if you want this feature.
++
+ config CHR_DEV_ST
+ tristate "SCSI tape support"
+ depends on SCSI
+Index: linux+rhel4+chaos/drivers/scsi/sd.c
+===================================================================
+--- linux+rhel4+chaos.orig/drivers/scsi/sd.c
++++ linux+rhel4+chaos/drivers/scsi/sd.c
+@@ -63,6 +63,38 @@
+
+ #include "scsi_logging.h"
+
++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS))
++# include <linux/proc_fs.h>
++# include <linux/seq_file.h>
++
++typedef struct {
++ unsigned long long iostat_size;
++ unsigned long long iostat_count;
++} iostat_counter_t;
++
++#define IOSTAT_NCOUNTERS 16
++typedef struct {
++ iostat_counter_t iostat_read_histogram[IOSTAT_NCOUNTERS];
++ iostat_counter_t iostat_write_histogram[IOSTAT_NCOUNTERS];
++ struct timeval iostat_timeval;
++} iostat_stats_t;
++
++iostat_stats_t **sd_iostats;
++spinlock_t sd_iostats_lock;
++struct proc_dir_entry *sd_iostats_procdir;
++char sd_iostats_procdir_name[] = "sd_iostats";
++
++extern void sd_iostats_init(void);
++extern void sd_iostats_init_disk(struct gendisk *);
++extern void sd_iostats_fini(void);
++extern void sd_iostats_bump(int disk, unsigned int nsect, int iswrite);
++#else
++static inline void sd_iostats_init(void) {}
++static inline void sd_iostats_init_disk(struct gendisk *disk) {}
++static inline void sd_iostats_fini(void) {}
++static inline void sd_iostats_bump(int disk, unsigned int nsect, int iswrite) {}
++#endif
++
+ MODULE_AUTHOR("Eric Youngdale");
+ MODULE_DESCRIPTION("SCSI disk (sd) driver");
+ MODULE_LICENSE("GPL");
+@@ -89,6 +121,7 @@ MODULE_ALIAS_SCSI_DEVICE(TYPE_RBC);
+ static DEFINE_IDR(sd_index_idr);
+ static DEFINE_SPINLOCK(sd_index_lock);
+
++#define SD_STATS 256
+ /* This semaphore is used to mediate the 0->1 reference get in the
+ * face of object destruction (i.e. we can't allow a get on an
+ * object after last put) */
+@@ -368,6 +401,9 @@ static int sd_init_command(struct scsi_c
+ SCSI_LOG_HLQUEUE(2, scmd_printk(KERN_INFO, SCpnt, "block=%llu\n",
+ (unsigned long long)block));
+
++ sd_iostats_bump(scsi_disk(disk)->index, this_count,
++ rq_data_dir(SCpnt->request) == WRITE);
++
+ /*
+ * If we have a 1K hardware sectorsize, prevent access to single
+ * 512 byte sectors. In theory we could handle this - in fact
+@@ -575,6 +611,7 @@ static int sd_open(struct inode *inode,
+ scsi_set_medium_removal(sdev, SCSI_REMOVAL_PREVENT);
+ }
+
++ sd_iostats_init_disk(disk);
+ return 0;
+
+ error_out:
+@@ -601,8 +638,20 @@ static int sd_release(struct inode *inod
+
+ SCSI_LOG_HLQUEUE(3, sd_printk(KERN_INFO, sdkp, "sd_release\n"));
+
+- if (!--sdkp->openers && sdev->removable) {
+- if (scsi_block_when_processing_errors(sdev))
++ if (!--sdkp->openers) {
++ /*
++ * Remove sd_iostats information about this disk
++ */
++ if (sd_iostats_procdir != NULL) {
++ remove_proc_entry(disk->disk_name, sd_iostats_procdir);
++ }
++ if (sd_iostats != NULL) {
++ if (sd_iostats[sdkp->index] != NULL) {
++ kfree (sd_iostats[sdkp->index]);
++ sd_iostats[sdkp->index] = NULL;
++ }
++ }
++ if (sdev->removable && scsi_block_when_processing_errors(sdev))
+ scsi_set_medium_removal(sdev, SCSI_REMOVAL_ALLOW);
+ }
+
+@@ -1563,6 +1612,342 @@ static int sd_revalidate_disk(struct gen
+ return 0;
+ }
+
++#if (defined(CONFIG_SD_IOSTATS) && defined(CONFIG_PROC_FS))
++static int
++sd_iostats_seq_show(struct seq_file *seq, void *v)
++{
++ struct timeval now;
++ struct gendisk *disk;
++ iostat_stats_t *stats;
++ unsigned long long read_len;
++ unsigned long long read_len_tot;
++ unsigned long read_num;
++ unsigned long read_num_tot;
++ unsigned long long write_len;
++ unsigned long long write_len_tot;
++ unsigned long write_num;
++ unsigned long write_num_tot;
++ int i;
++ int maxi;
++
++ if (seq == NULL || seq->private == NULL) {
++ printk(KERN_ERR "sd_iostats_seq_show: NULL disk\n");
++ BUG();
++ }
++
++ disk = seq->private;
++
++ if (scsi_disk(disk) == NULL || (disk->flags & GENHD_FL_UP) == 0) {
++ seq_printf(seq, "sd_iostats_seq_show: Device %s "
++ "does not exist\n", disk->disk_name);
++ return 0;
++ }
++
++ if (sd_iostats == NULL) {
++ printk(KERN_ERR "sd_iostats_seq_show: NULL stats array\n");
++ BUG();
++ }
++
++ stats = sd_iostats[scsi_disk(disk)->index];
++ if (stats == NULL) {
++ seq_printf(seq, "sd_iostats_seq_show: sd_iostats "
++ "entry %d does not exist\n",
++ scsi_disk(disk)->index);
++ return 0;
++ }
++
++ do_gettimeofday(&now);
++ now.tv_sec -= stats->iostat_timeval.tv_sec;
++ now.tv_usec -= stats->iostat_timeval.tv_usec;
++ if (now.tv_usec < 0) {
++ now.tv_usec += 1000000;
++ now.tv_sec--;
++ }
++
++ /* this sampling races with updates */
++ seq_printf(seq, "index: %lu snapshot_time: %lu.%06lu\n",
++ scsi_disk(disk)->index, now.tv_sec, now.tv_usec);
++
++ for (i = IOSTAT_NCOUNTERS - 1; i > 0; i--)
++ if (stats->iostat_read_histogram[i].iostat_count != 0 ||
++ stats->iostat_write_histogram[i].iostat_count != 0)
++ break;
++ maxi = i;
++
++ seq_printf(seq, "%8s %8s %12s %8s %12s\n", "size",
++ "reads", "total", "writes", "total");
++
++ read_len_tot = write_len_tot = 0;
++ read_num_tot = write_num_tot = 0;
++ for (i = 0; i <= maxi; i++) {
++ read_len = stats->iostat_read_histogram[i].iostat_size;
++ read_len_tot += read_len;
++ read_num = stats->iostat_read_histogram[i].iostat_count;
++ read_num_tot += read_num;
++
++ write_len = stats->iostat_write_histogram[i].iostat_size;
++ write_len_tot += write_len;
++ write_num = stats->iostat_write_histogram[i].iostat_count;
++ write_num_tot += write_num;
++
++ seq_printf (seq, "%8d %8lu %12llu %8lu %12llu\n",
++ 512<<i, read_num, read_len, write_num, write_len);
++ }
++
++ seq_printf(seq, "%8s %8lu %12llu %8lu %12llu\n", "total",
++ read_num_tot, read_len_tot,
++ write_num_tot, write_len_tot);
++ return 0;
++}
++
++static void *
++sd_iostats_seq_start(struct seq_file *p, loff_t *pos)
++{
++ return (*pos == 0) ? (void *)1 : NULL;
++}
++
++static void *
++sd_iostats_seq_next(struct seq_file *p, void *v, loff_t *pos)
++{
++ ++*pos;
++ return NULL;
++}
++
++static void
++sd_iostats_seq_stop(struct seq_file *p, void *v)
++{
++}
++
++static struct seq_operations sd_iostats_seqops = {
++ .start = sd_iostats_seq_start,
++ .stop = sd_iostats_seq_stop,
++ .next = sd_iostats_seq_next,
++ .show = sd_iostats_seq_show,
++};
++
++static int
++sd_iostats_seq_open (struct inode *inode, struct file *file)
++{
++ int rc;
++
++ rc = seq_open(file, &sd_iostats_seqops);
++ if (rc != 0)
++ return rc;
++
++ ((struct seq_file *)file->private_data)->private = PDE(inode)->data;
++ return 0;
++}
++
++static ssize_t
++sd_iostats_seq_write(struct file *file, const char *buffer,
++ size_t len, loff_t *off)
++{
++ struct seq_file *seq = file->private_data;
++ struct gendisk *disk = seq->private;
++ iostat_stats_t *stats = sd_iostats[scsi_disk(disk)->index];
++ unsigned long flags;
++
++
++ spin_lock_irqsave (&sd_iostats_lock, flags);
++ memset (stats, 0, sizeof(*stats));
++ do_gettimeofday(&stats->iostat_timeval);
++ spin_unlock_irqrestore (&sd_iostats_lock, flags);
++
++ return len;
++}
++
++static struct file_operations sd_iostats_proc_fops = {
++ .owner = THIS_MODULE,
++ .open = sd_iostats_seq_open,
++ .read = seq_read,
++ .write = sd_iostats_seq_write,
++ .llseek = seq_lseek,
++ .release = seq_release,
++};
++
++extern struct proc_dir_entry *proc_scsi;
++
++void
++sd_iostats_init(void)
++{
++ int i;
++
++ spin_lock_init(&sd_iostats_lock);
++
++ sd_iostats = kmalloc(SD_STATS * sizeof(iostat_stats_t *), GFP_KERNEL);
++ if (sd_iostats == NULL) {
++ printk(KERN_WARNING "Can't keep sd iostats: "
++ "ENOMEM allocating stats array size %ld\n",
++ SD_STATS * sizeof(iostat_stats_t *));
++ return;
++ }
++
++ for (i = 0; i < SD_STATS; i++)
++ sd_iostats[i] = NULL;
++
++ if (proc_scsi == NULL) {
++ printk(KERN_WARNING "No access to sd iostats: "
++ "proc_scsi is NULL\n");
++ return;
++ }
++
++ sd_iostats_procdir = create_proc_entry(sd_iostats_procdir_name,
++ S_IFDIR | S_IRUGO | S_IXUGO,
++ proc_scsi);
++ if (sd_iostats_procdir == NULL) {
++ printk(KERN_WARNING "No access to sd iostats: "
++ "can't create /proc/scsi/%s\n", sd_iostats_procdir_name);
++ return;
++ }
++}
++
++void
++sd_iostats_init_disk(struct gendisk *disk)
++{
++ struct proc_dir_entry *pde;
++ unsigned long flags;
++ iostat_stats_t *stats;
++
++ if (sd_iostats == NULL ||
++ sd_iostats_procdir == NULL)
++ return;
++
++ if (scsi_disk(disk)->index > SD_STATS) {
++ printk(KERN_ERR "sd_iostats_init_disk: "
++ "unexpected disk index %d(%d)\n",
++ scsi_disk(disk)->index, SD_STATS);
++ return;
++ }
++
++ if (sd_iostats[scsi_disk(disk)->index] != NULL)
++ return;
++
++ stats = kmalloc(sizeof(*stats), GFP_KERNEL);
++ if (stats == NULL) {
++ printk(KERN_WARNING "Can't keep %s iostats: "
++ "ENOMEM allocating stats size %ld\n",
++ disk->disk_name, sizeof(*stats));
++ return;
++ }
++
++ memset (stats, 0, sizeof(*stats));
++ do_gettimeofday(&stats->iostat_timeval);
++
++ spin_lock_irqsave(&sd_iostats_lock, flags);
++
++ if (sd_iostats[scsi_disk(disk)->index] != NULL) {
++ spin_unlock_irqrestore(&sd_iostats_lock, flags);
++ kfree (stats);
++ return;
++ }
++
++ sd_iostats[scsi_disk(disk)->index] = stats;
++
++ spin_unlock_irqrestore(&sd_iostats_lock, flags);
++
++ pde = create_proc_entry(disk->disk_name, S_IRUGO | S_IWUSR,
++ sd_iostats_procdir);
++ if (pde == NULL) {
++ printk(KERN_WARNING "Can't create /proc/scsi/%s/%s\n",
++ sd_iostats_procdir_name, disk->disk_name);
++ } else {
++ pde->proc_fops = &sd_iostats_proc_fops;
++ pde->data = disk;
++ }
++}
++
++static void sd_devname(unsigned int disknum, char *buffer)
++{
++ if (disknum < 26)
++ sprintf(buffer, "sd%c", 'a' + disknum);
++ else {
++ unsigned int min1;
++ unsigned int min2;
++ /*
++ * For larger numbers of disks, we need to go to a new
++ * naming scheme.
++ */
++ min1 = disknum / 26;
++ min2 = disknum % 26;
++ sprintf(buffer, "sd%c%c", 'a' + min1 - 1, 'a' + min2);
++ }
++}
++
++void
++sd_iostats_fini(void)
++{
++ char name[6];
++ int i;
++
++ if (sd_iostats_procdir != NULL) {
++ for (i = 0; i < SD_STATS; i++) {
++ sd_devname(i, name);
++ remove_proc_entry(name, sd_iostats_procdir);
++ }
++
++ if (proc_scsi == NULL) {
++ printk(KERN_ERR "sd_iostats_fini: proc_scsi NULL\n");
++ BUG();
++ }
++ remove_proc_entry(sd_iostats_procdir_name,
++ proc_scsi);
++
++ sd_iostats_procdir = NULL;
++ }
++
++ if (sd_iostats != NULL) {
++ for (i = 0; i < SD_STATS; i++) {
++ if (sd_iostats[i] != NULL)
++ kfree (sd_iostats[i]);
++ }
++
++ kfree(sd_iostats);
++ sd_iostats = NULL;
++ }
++}
++
++void
++sd_iostats_bump(int disk, unsigned int nsect, int iswrite)
++{
++ iostat_stats_t *stats;
++ iostat_counter_t *counter;
++ int bucket;
++ int tmp;
++ unsigned long irqflags;
++
++ if (sd_iostats == NULL)
++ return;
++
++ if (disk < 0 || disk >= SD_STATS) {
++ printk(KERN_ERR "sd_iostats_bump: unexpected disk index %d([0-%d])\n",
++ disk, SD_STATS);
++ BUG();
++ }
++
++ for (bucket = 0, tmp = nsect; tmp > 1; bucket++)
++ tmp /= 2;
++
++ if (bucket >= IOSTAT_NCOUNTERS) {
++ printk (KERN_ERR "sd_iostats_bump: nsect %d too big\n", nsect);
++ BUG();
++ }
++
++ spin_lock_irqsave(&sd_iostats_lock, irqflags);
++
++ stats = sd_iostats[disk];
++ if (stats != NULL) {
++ counter = iswrite ?
++ &stats->iostat_write_histogram[bucket] :
++ &stats->iostat_read_histogram[bucket];
++
++ counter->iostat_size += nsect;
++ counter->iostat_count++;
++ }
++
++ spin_unlock_irqrestore(&sd_iostats_lock, irqflags);
++}
++#endif
++
+ /**
+ * sd_probe - called during driver initialization and whenever a
+ * new scsi device is attached to the system. It is called once
+@@ -1854,6 +2239,7 @@ static int __init init_sd(void)
+ err = scsi_register_driver(&sd_template.gendrv);
+ if (err)
+ goto err_out_class;
++ sd_iostats_init();
+
+ return 0;
+
+@@ -1876,6 +2262,7 @@ static void __exit exit_sd(void)
+
+ SCSI_LOG_HLQUEUE(3, printk("exit_sd: exiting sd driver\n"));
+
++ sd_iostats_fini();
+ scsi_unregister_driver(&sd_template.gendrv);
+ class_unregister(&sd_disk_class);
+
+Index: linux-2.6.22.5/drivers/scsi/scsi_proc.c
+===================================================================
+--- linux-2.6.22.5.orig/drivers/scsi/scsi_proc.c
++++ linux-2.6.22.5/drivers/scsi/scsi_proc.c
+@@ -40,7 +40,8 @@
+ /* 4K page size, but our output routines, use some slack for overruns */
+ #define PROC_BLOCK_SIZE (3*1024)
+
+-static struct proc_dir_entry *proc_scsi;
++struct proc_dir_entry *proc_scsi;
++EXPORT_SYMBOL(proc_scsi);
+
+ /* Protect sht->present and sht->proc_dir */
+ static DEFINE_MUTEX(global_host_template_mutex);
git://git.whamcloud.com / fs / lustre-release.git / commitdiff