> > Posting the information about why the problem was being seen is both ontopic
> > and useful. I've been able to cross several mandrake bug reports off my weird
> > kernel bug list knowing this (and to point the submitters at the update)
>
> And I'm just experiencing againg a rollback segment corruption again, and it
> seems there hasn't been any crash. So I suspect there is something else :-(
>
> As you suggested, I'll try to upgrade them to a 2.2.13 compiled with a egcs.
> I noticed a similarity with two sites on which I had corruption: they both
> running HP servers with a HP NetRaid (amird) controller. One runs a 2.0.36
> kernel, and the other a 2.2.9 kernel (RedHat 5.2 and Mandrake 6.0).
>
> Maybe it could be something wrong with amird driver or HP NetRaid firmware ?
I guess it is the time...
Here it is: module into kernel to support adding crc's. Then, broken
firmware will be non-issue.
Relative to 2.3.latest, and you'll need some userland support even I
do not have, yet. Not much tested, but it worked on small system. I
would not run it at production system if I were you.
This needs someone with flakey system to work on it.
Pavel
--- clean/drivers/block/loop.c Sun Nov 7 22:20:37 1999
+++ linux/drivers/block/loop.c Fri Nov 19 23:04:38 1999
@@ -54,6 +54,7 @@
#include <linux/init.h>
#include <asm/uaccess.h>
+#include <asm/checksum.h>
#include <linux/loop.h>
@@ -114,6 +115,97 @@
return 0;
}
+#define ID printk(KERN_ERR "crc: info about (%s, %d, %d) ", kdevname(lo->lo_device), real_block, blksize);
+ /* CRC method copyright 1999 Pavel Machek <pavel@suse.cz> */
+
+static int transfer_crc(struct loop_device *lo, int cmd, char *raw_buf,
+ char *loop_buf, int size, int real_block)
+{
+ struct buffer_head *bh;
+ int blksize = 4096, nsect; /* Size of block on auxilary media */
+ int cksum;
+ u32 *data;
+ nsect = blksize / 4;
+
+ bh = getblk(lo->second_device, 1+real_block/nsect, blksize);
+ if (!bh) {
+ ID; printk( "getblk returned NULL.\n" );
+ return -1;
+ }
+ if (!buffer_uptodate(bh)) {
+ ll_rw_block(READ, 1, &bh);
+ wait_on_buffer(bh);
+ if (!buffer_uptodate(bh)) {
+ ID; printk( "could not read block with CRC\n" );
+ goto error;
+ }
+ }
+
+ data = bh->b_data;
+ if (cmd == READ)
+ cksum = csum_partial_copy_nocheck(raw_buf, loop_buf, size, 0);
+ else
+ cksum = csum_partial_copy_nocheck(loop_buf, raw_buf, size, 0);
+
+ if (cmd == READ) {
+ if (le32_to_cpu(data[real_block%nsect]) != cksum) {
+ if (lo->lo_encrypt_key_size == 0) { /* Normal mode */
+ ID; printk( "wrong checksum reading, is %x, should be %x\n", cksum, 0x1234 );
+ goto error;
+ } else {
+ ID; printk( "wrong checksum repairing, setting to %x\n", cksum );
+ goto repair;
+ }
+ }
+ } else {
+ repair:
+ data[real_block%nsect] = cpu_to_le32(cksum);
+ mark_buffer_uptodate(bh, 1);
+ mark_buffer_dirty(bh, 1);
+ }
+
+ brelse(bh);
+ return 0;
+error:
+ brelse(bh);
+ return -1;
+
+}
+
+static int ioctl_crc(struct loop_device *lo, int cmd, unsigned long arg)
+{
+ struct file *file;
+ struct inode *inode;
+ int error;
+
+ printk( "Entering ioctl_crc\n" );
+ if (cmd != LOOP_CRC_SET_FD)
+ return -EINVAL;
+
+ error = -EBADF;
+ file = fget(arg);
+ if (!file)
+ return -EINVAL;
+
+ error = -EINVAL;
+ inode = file->f_dentry->d_inode;
+ if (!inode) {
+ printk(KERN_ERR "ioctl_crc: NULL inode?!?\n");
+ goto out;
+ }
+
+ if (S_ISBLK(inode->i_mode)) {
+ error = blkdev_open(inode, file);
+ lo->second_device = inode->i_rdev;
+ printk( "loop_crc: Registered device %x\n", lo->second_device );
+ return error;
+ } else {
+ out:
+ fput(file);
+ return -EINVAL;
+ }
+}
+
static int none_status(struct loop_device *lo, struct loop_info *info)
{
return 0;
@@ -138,10 +230,19 @@
init: xor_status
};
+struct loop_func_table crc_funcs = {
+ number: LO_CRYPT_CRC,
+ transfer: transfer_crc,
+ init: none_status,
+ ioctl: ioctl_crc
+};
+
/* xfer_funcs[0] is special - its release function is never called */
struct loop_func_table *xfer_funcs[MAX_LO_CRYPT] = {
&none_funcs,
- &xor_funcs
+ &xor_funcs,
+ NULL, NULL, NULL, NULL, NULL,
+ &crc_funcs,
};
#define MAX_DISK_SIZE 1024*1024*1024
@@ -496,6 +597,7 @@
lo->transfer = NULL;
lo->ioctl = NULL;
lo->lo_device = 0;
+ lo->second_device = 0;
lo->lo_encrypt_type = 0;
lo->lo_offset = 0;
lo->lo_encrypt_key_size = 0;
--- clean/include/linux/loop.h Tue Dec 1 22:11:49 1998
+++ linux/include/linux/loop.h Fri Nov 19 23:04:40 1999
@@ -22,6 +22,7 @@
struct dentry *lo_dentry;
int lo_refcnt;
kdev_t lo_device;
+ kdev_t second_device; /* For crc method, device to write crc's to */
int lo_offset;
int lo_encrypt_type;
int lo_encrypt_key_size;
@@ -58,9 +59,11 @@
* from a glibc program, because glibc has a 32bit dev_t.
* Prevent people from shooting in their own foot.
*/
-#if __GLIBC__ >= 2 && !defined(dev_t)
+#if 0
+/* __GLIBC__ >= 2 && !defined(dev_t) */
#error "Wrong dev_t in loop.h"
#endif
+#define dev_t __u16
/*
* This uses kdev_t because glibc currently has no appropiate
@@ -94,6 +97,7 @@
#define LO_CRYPT_BLOW 4
#define LO_CRYPT_CAST128 5
#define LO_CRYPT_IDEA 6
+#define LO_CRYPT_CRC 7
#define LO_CRYPT_DUMMY 9
#define LO_CRYPT_SKIPJACK 10
#define MAX_LO_CRYPT 20
@@ -126,5 +130,6 @@
#define LOOP_CLR_FD 0x4C01
#define LOOP_SET_STATUS 0x4C02
#define LOOP_GET_STATUS 0x4C03
+#define LOOP_CRC_SET_FD 0x4C04
#endif
-- I'm really pavel@ucw.cz. Look at http://195.113.31.123/~pavel. Pavel Hi! I'm a .signature virus! Copy me into your ~/.signature, please!- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/