linux/drivers/block/nbd.c

739 lines
19 KiB
C
Raw Normal View History

/*
* Network block device - make block devices work over TCP
*
* Note that you can not swap over this thing, yet. Seems to work but
* deadlocks sometimes - you can not swap over TCP in general.
*
* Copyright 1997-2000 Pavel Machek <pavel@ucw.cz>
* Parts copyright 2001 Steven Whitehouse <steve@chygwyn.com>
*
* (part of code stolen from loop.c)
*
* 97-3-25 compiled 0-th version, not yet tested it
* (it did not work, BTW) (later that day) HEY! it works!
* (bit later) hmm, not that much... 2:00am next day:
* yes, it works, but it gives something like 50kB/sec
* 97-4-01 complete rewrite to make it possible for many requests at
* once to be processed
* 97-4-11 Making protocol independent of endianity etc.
* 97-9-13 Cosmetic changes
* 98-5-13 Attempt to make 64-bit-clean on 64-bit machines
* 99-1-11 Attempt to make 64-bit-clean on 32-bit machines <ankry@mif.pg.gda.pl>
* 01-2-27 Fix to store proper blockcount for kernel (calculated using
* BLOCK_SIZE_BITS, not device blocksize) <aga@permonline.ru>
* 01-3-11 Make nbd work with new Linux block layer code. It now supports
* plugging like all the other block devices. Also added in MSG_MORE to
* reduce number of partial TCP segments sent. <steve@chygwyn.com>
* 01-12-6 Fix deadlock condition by making queue locks independent of
* the transmit lock. <steve@chygwyn.com>
* 02-10-11 Allow hung xmit to be aborted via SIGKILL & various fixes.
* <Paul.Clements@SteelEye.com> <James.Bottomley@SteelEye.com>
* 03-06-22 Make nbd work with new linux 2.5 block layer design. This fixes
* memory corruption from module removal and possible memory corruption
* from sending/receiving disk data. <ldl@aros.net>
* 03-06-23 Cosmetic changes. <ldl@aros.net>
* 03-06-23 Enhance diagnostics support. <ldl@aros.net>
* 03-06-24 Remove unneeded blksize_bits field from nbd_device struct.
* <ldl@aros.net>
* 03-06-24 Cleanup PARANOIA usage & code. <ldl@aros.net>
* 04-02-19 Remove PARANOIA, plus various cleanups (Paul Clements)
* possible FIXME: make set_sock / set_blksize / set_size / do_it one syscall
* why not: would need access_ok and friends, would share yet another
* structure with userland
*/
#include <linux/major.h>
#include <linux/blkdev.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/fs.h>
#include <linux/bio.h>
#include <linux/stat.h>
#include <linux/errno.h>
#include <linux/file.h>
#include <linux/ioctl.h>
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
#include <linux/compiler.h>
#include <linux/err.h>
#include <linux/kernel.h>
#include <net/sock.h>
#include <linux/devfs_fs_kernel.h>
#include <asm/uaccess.h>
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
#include <asm/system.h>
#include <asm/types.h>
#include <linux/nbd.h>
#define LO_MAGIC 0x68797548
#ifdef NDEBUG
#define dprintk(flags, fmt...)
#else /* NDEBUG */
#define dprintk(flags, fmt...) do { \
if (debugflags & (flags)) printk(KERN_DEBUG fmt); \
} while (0)
#define DBG_IOCTL 0x0004
#define DBG_INIT 0x0010
#define DBG_EXIT 0x0020
#define DBG_BLKDEV 0x0100
#define DBG_RX 0x0200
#define DBG_TX 0x0400
static unsigned int debugflags;
static unsigned int nbds_max = 16;
#endif /* NDEBUG */
static struct nbd_device nbd_dev[MAX_NBD];
/*
* Use just one lock (or at most 1 per NIC). Two arguments for this:
* 1. Each NIC is essentially a synchronization point for all servers
* accessed through that NIC so there's no need to have more locks
* than NICs anyway.
* 2. More locks lead to more "Dirty cache line bouncing" which will slow
* down each lock to the point where they're actually slower than just
* a single lock.
* Thanks go to Jens Axboe and Al Viro for their LKML emails explaining this!
*/
static DEFINE_SPINLOCK(nbd_lock);
#ifndef NDEBUG
static const char *ioctl_cmd_to_ascii(int cmd)
{
switch (cmd) {
case NBD_SET_SOCK: return "set-sock";
case NBD_SET_BLKSIZE: return "set-blksize";
case NBD_SET_SIZE: return "set-size";
case NBD_DO_IT: return "do-it";
case NBD_CLEAR_SOCK: return "clear-sock";
case NBD_CLEAR_QUE: return "clear-que";
case NBD_PRINT_DEBUG: return "print-debug";
case NBD_SET_SIZE_BLOCKS: return "set-size-blocks";
case NBD_DISCONNECT: return "disconnect";
case BLKROSET: return "set-read-only";
case BLKFLSBUF: return "flush-buffer-cache";
}
return "unknown";
}
static const char *nbdcmd_to_ascii(int cmd)
{
switch (cmd) {
case NBD_CMD_READ: return "read";
case NBD_CMD_WRITE: return "write";
case NBD_CMD_DISC: return "disconnect";
}
return "invalid";
}
#endif /* NDEBUG */
static void nbd_end_request(struct request *req)
{
int uptodate = (req->errors == 0) ? 1 : 0;
request_queue_t *q = req->q;
unsigned long flags;
dprintk(DBG_BLKDEV, "%s: request %p: %s\n", req->rq_disk->disk_name,
req, uptodate? "done": "failed");
spin_lock_irqsave(q->queue_lock, flags);
if (!end_that_request_first(req, uptodate, req->nr_sectors)) {
end_that_request_last(req, uptodate);
}
spin_unlock_irqrestore(q->queue_lock, flags);
}
/*
* Send or receive packet.
*/
static int sock_xmit(struct socket *sock, int send, void *buf, int size,
int msg_flags)
{
int result;
struct msghdr msg;
struct kvec iov;
unsigned long flags;
sigset_t oldset;
/* Allow interception of SIGKILL only
* Don't allow other signals to interrupt the transmission */
spin_lock_irqsave(&current->sighand->siglock, flags);
oldset = current->blocked;
sigfillset(&current->blocked);
sigdelsetmask(&current->blocked, sigmask(SIGKILL));
recalc_sigpending();
spin_unlock_irqrestore(&current->sighand->siglock, flags);
do {
sock->sk->sk_allocation = GFP_NOIO;
iov.iov_base = buf;
iov.iov_len = size;
msg.msg_name = NULL;
msg.msg_namelen = 0;
msg.msg_control = NULL;
msg.msg_controllen = 0;
msg.msg_flags = msg_flags | MSG_NOSIGNAL;
if (send)
result = kernel_sendmsg(sock, &msg, &iov, 1, size);
else
result = kernel_recvmsg(sock, &msg, &iov, 1, size, 0);
if (signal_pending(current)) {
siginfo_t info;
spin_lock_irqsave(&current->sighand->siglock, flags);
printk(KERN_WARNING "nbd (pid %d: %s) got signal %d\n",
current->pid, current->comm,
dequeue_signal(current, &current->blocked, &info));
spin_unlock_irqrestore(&current->sighand->siglock, flags);
result = -EINTR;
break;
}
if (result <= 0) {
if (result == 0)
result = -EPIPE; /* short read */
break;
}
size -= result;
buf += result;
} while (size > 0);
spin_lock_irqsave(&current->sighand->siglock, flags);
current->blocked = oldset;
recalc_sigpending();
spin_unlock_irqrestore(&current->sighand->siglock, flags);
return result;
}
static inline int sock_send_bvec(struct socket *sock, struct bio_vec *bvec,
int flags)
{
int result;
void *kaddr = kmap(bvec->bv_page);
result = sock_xmit(sock, 1, kaddr + bvec->bv_offset, bvec->bv_len,
flags);
kunmap(bvec->bv_page);
return result;
}
static int nbd_send_req(struct nbd_device *lo, struct request *req)
{
int result, i, flags;
struct nbd_request request;
unsigned long size = req->nr_sectors << 9;
struct socket *sock = lo->sock;
request.magic = htonl(NBD_REQUEST_MAGIC);
request.type = htonl(nbd_cmd(req));
request.from = cpu_to_be64((u64) req->sector << 9);
request.len = htonl(size);
memcpy(request.handle, &req, sizeof(req));
dprintk(DBG_TX, "%s: request %p: sending control (%s@%llu,%luB)\n",
lo->disk->disk_name, req,
nbdcmd_to_ascii(nbd_cmd(req)),
(unsigned long long)req->sector << 9,
req->nr_sectors << 9);
result = sock_xmit(sock, 1, &request, sizeof(request),
(nbd_cmd(req) == NBD_CMD_WRITE)? MSG_MORE: 0);
if (result <= 0) {
printk(KERN_ERR "%s: Send control failed (result %d)\n",
lo->disk->disk_name, result);
goto error_out;
}
if (nbd_cmd(req) == NBD_CMD_WRITE) {
struct bio *bio;
/*
* we are really probing at internals to determine
* whether to set MSG_MORE or not...
*/
rq_for_each_bio(bio, req) {
struct bio_vec *bvec;
bio_for_each_segment(bvec, bio, i) {
flags = 0;
if ((i < (bio->bi_vcnt - 1)) || bio->bi_next)
flags = MSG_MORE;
dprintk(DBG_TX, "%s: request %p: sending %d bytes data\n",
lo->disk->disk_name, req,
bvec->bv_len);
result = sock_send_bvec(sock, bvec, flags);
if (result <= 0) {
printk(KERN_ERR "%s: Send data failed (result %d)\n",
lo->disk->disk_name,
result);
goto error_out;
}
}
}
}
return 0;
error_out:
return 1;
}
static struct request *nbd_find_request(struct nbd_device *lo, char *handle)
{
struct request *req;
struct list_head *tmp;
struct request *xreq;
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
int err;
memcpy(&xreq, handle, sizeof(xreq));
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
err = wait_event_interruptible(lo->active_wq, lo->active_req != xreq);
if (unlikely(err))
goto out;
spin_lock(&lo->queue_lock);
list_for_each(tmp, &lo->queue_head) {
req = list_entry(tmp, struct request, queuelist);
if (req != xreq)
continue;
list_del_init(&req->queuelist);
spin_unlock(&lo->queue_lock);
return req;
}
spin_unlock(&lo->queue_lock);
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
err = -ENOENT;
out:
return ERR_PTR(err);
}
static inline int sock_recv_bvec(struct socket *sock, struct bio_vec *bvec)
{
int result;
void *kaddr = kmap(bvec->bv_page);
result = sock_xmit(sock, 0, kaddr + bvec->bv_offset, bvec->bv_len,
MSG_WAITALL);
kunmap(bvec->bv_page);
return result;
}
/* NULL returned = something went wrong, inform userspace */
static struct request *nbd_read_stat(struct nbd_device *lo)
{
int result;
struct nbd_reply reply;
struct request *req;
struct socket *sock = lo->sock;
reply.magic = 0;
result = sock_xmit(sock, 0, &reply, sizeof(reply), MSG_WAITALL);
if (result <= 0) {
printk(KERN_ERR "%s: Receive control failed (result %d)\n",
lo->disk->disk_name, result);
goto harderror;
}
req = nbd_find_request(lo, reply.handle);
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
if (unlikely(IS_ERR(req))) {
result = PTR_ERR(req);
if (result != -ENOENT)
goto harderror;
printk(KERN_ERR "%s: Unexpected reply (%p)\n",
lo->disk->disk_name, reply.handle);
result = -EBADR;
goto harderror;
}
if (ntohl(reply.magic) != NBD_REPLY_MAGIC) {
printk(KERN_ERR "%s: Wrong magic (0x%lx)\n",
lo->disk->disk_name,
(unsigned long)ntohl(reply.magic));
result = -EPROTO;
goto harderror;
}
if (ntohl(reply.error)) {
printk(KERN_ERR "%s: Other side returned error (%d)\n",
lo->disk->disk_name, ntohl(reply.error));
req->errors++;
return req;
}
dprintk(DBG_RX, "%s: request %p: got reply\n",
lo->disk->disk_name, req);
if (nbd_cmd(req) == NBD_CMD_READ) {
int i;
struct bio *bio;
rq_for_each_bio(bio, req) {
struct bio_vec *bvec;
bio_for_each_segment(bvec, bio, i) {
result = sock_recv_bvec(sock, bvec);
if (result <= 0) {
printk(KERN_ERR "%s: Receive data failed (result %d)\n",
lo->disk->disk_name,
result);
goto harderror;
}
dprintk(DBG_RX, "%s: request %p: got %d bytes data\n",
lo->disk->disk_name, req, bvec->bv_len);
}
}
}
return req;
harderror:
lo->harderror = result;
return NULL;
}
static void nbd_do_it(struct nbd_device *lo)
{
struct request *req;
BUG_ON(lo->magic != LO_MAGIC);
while ((req = nbd_read_stat(lo)) != NULL)
nbd_end_request(req);
return;
}
static void nbd_clear_que(struct nbd_device *lo)
{
struct request *req;
BUG_ON(lo->magic != LO_MAGIC);
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
/*
* Because we have set lo->sock to NULL under the tx_lock, all
* modifications to the list must have completed by now. For
* the same reason, the active_req must be NULL.
*
* As a consequence, we don't need to take the spin lock while
* purging the list here.
*/
BUG_ON(lo->sock);
BUG_ON(lo->active_req);
while (!list_empty(&lo->queue_head)) {
req = list_entry(lo->queue_head.next, struct request,
queuelist);
list_del_init(&req->queuelist);
req->errors++;
nbd_end_request(req);
}
}
/*
* We always wait for result of write, for now. It would be nice to make it optional
* in future
* if ((req->cmd == WRITE) && (lo->flags & NBD_WRITE_NOCHK))
* { printk( "Warning: Ignoring result!\n"); nbd_end_request( req ); }
*/
static void do_nbd_request(request_queue_t * q)
{
struct request *req;
while ((req = elv_next_request(q)) != NULL) {
struct nbd_device *lo;
blkdev_dequeue_request(req);
dprintk(DBG_BLKDEV, "%s: request %p: dequeued (flags=%lx)\n",
req->rq_disk->disk_name, req, req->flags);
if (!(req->flags & REQ_CMD))
goto error_out;
lo = req->rq_disk->private_data;
BUG_ON(lo->magic != LO_MAGIC);
nbd_cmd(req) = NBD_CMD_READ;
if (rq_data_dir(req) == WRITE) {
nbd_cmd(req) = NBD_CMD_WRITE;
if (lo->flags & NBD_READ_ONLY) {
printk(KERN_ERR "%s: Write on read-only\n",
lo->disk->disk_name);
goto error_out;
}
}
req->errors = 0;
spin_unlock_irq(q->queue_lock);
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
down(&lo->tx_lock);
if (unlikely(!lo->sock)) {
up(&lo->tx_lock);
printk(KERN_ERR "%s: Attempted send on closed socket\n",
lo->disk->disk_name);
req->errors++;
nbd_end_request(req);
spin_lock_irq(q->queue_lock);
continue;
}
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
lo->active_req = req;
if (nbd_send_req(lo, req) != 0) {
printk(KERN_ERR "%s: Request send failed\n",
lo->disk->disk_name);
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
req->errors++;
nbd_end_request(req);
} else {
spin_lock(&lo->queue_lock);
list_add(&req->queuelist, &lo->queue_head);
spin_unlock(&lo->queue_lock);
}
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
lo->active_req = NULL;
up(&lo->tx_lock);
wake_up_all(&lo->active_wq);
spin_lock_irq(q->queue_lock);
continue;
error_out:
req->errors++;
spin_unlock(q->queue_lock);
nbd_end_request(req);
spin_lock(q->queue_lock);
}
return;
}
static int nbd_ioctl(struct inode *inode, struct file *file,
unsigned int cmd, unsigned long arg)
{
struct nbd_device *lo = inode->i_bdev->bd_disk->private_data;
int error;
struct request sreq ;
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
BUG_ON(lo->magic != LO_MAGIC);
/* Anyone capable of this syscall can do *real bad* things */
dprintk(DBG_IOCTL, "%s: nbd_ioctl cmd=%s(0x%x) arg=%lu\n",
lo->disk->disk_name, ioctl_cmd_to_ascii(cmd), cmd, arg);
switch (cmd) {
case NBD_DISCONNECT:
printk(KERN_INFO "%s: NBD_DISCONNECT\n", lo->disk->disk_name);
sreq.flags = REQ_SPECIAL;
nbd_cmd(&sreq) = NBD_CMD_DISC;
/*
* Set these to sane values in case server implementation
* fails to check the request type first and also to keep
* debugging output cleaner.
*/
sreq.sector = 0;
sreq.nr_sectors = 0;
if (!lo->sock)
return -EINVAL;
nbd_send_req(lo, &sreq);
return 0;
case NBD_CLEAR_SOCK:
error = 0;
down(&lo->tx_lock);
lo->sock = NULL;
up(&lo->tx_lock);
file = lo->file;
lo->file = NULL;
nbd_clear_que(lo);
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
BUG_ON(!list_empty(&lo->queue_head));
if (file)
fput(file);
return error;
case NBD_SET_SOCK:
if (lo->file)
return -EBUSY;
error = -EINVAL;
file = fget(arg);
if (file) {
inode = file->f_dentry->d_inode;
if (S_ISSOCK(inode->i_mode)) {
lo->file = file;
lo->sock = SOCKET_I(inode);
error = 0;
} else {
fput(file);
}
}
return error;
case NBD_SET_BLKSIZE:
lo->blksize = arg;
lo->bytesize &= ~(lo->blksize-1);
inode->i_bdev->bd_inode->i_size = lo->bytesize;
set_blocksize(inode->i_bdev, lo->blksize);
set_capacity(lo->disk, lo->bytesize >> 9);
return 0;
case NBD_SET_SIZE:
lo->bytesize = arg & ~(lo->blksize-1);
inode->i_bdev->bd_inode->i_size = lo->bytesize;
set_blocksize(inode->i_bdev, lo->blksize);
set_capacity(lo->disk, lo->bytesize >> 9);
return 0;
case NBD_SET_SIZE_BLOCKS:
lo->bytesize = ((u64) arg) * lo->blksize;
inode->i_bdev->bd_inode->i_size = lo->bytesize;
set_blocksize(inode->i_bdev, lo->blksize);
set_capacity(lo->disk, lo->bytesize >> 9);
return 0;
case NBD_DO_IT:
if (!lo->file)
return -EINVAL;
nbd_do_it(lo);
/* on return tidy up in case we have a signal */
/* Forcibly shutdown the socket causing all listeners
* to error
*
* FIXME: This code is duplicated from sys_shutdown, but
* there should be a more generic interface rather than
* calling socket ops directly here */
down(&lo->tx_lock);
if (lo->sock) {
printk(KERN_WARNING "%s: shutting down socket\n",
lo->disk->disk_name);
lo->sock->ops->shutdown(lo->sock,
SEND_SHUTDOWN|RCV_SHUTDOWN);
lo->sock = NULL;
}
up(&lo->tx_lock);
file = lo->file;
lo->file = NULL;
nbd_clear_que(lo);
printk(KERN_WARNING "%s: queue cleared\n", lo->disk->disk_name);
if (file)
fput(file);
return lo->harderror;
case NBD_CLEAR_QUE:
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
/*
* This is for compatibility only. The queue is always cleared
* by NBD_DO_IT or NBD_CLEAR_SOCK.
*/
BUG_ON(!lo->sock && !list_empty(&lo->queue_head));
return 0;
case NBD_PRINT_DEBUG:
printk(KERN_INFO "%s: next = %p, prev = %p, head = %p\n",
inode->i_bdev->bd_disk->disk_name,
lo->queue_head.next, lo->queue_head.prev,
&lo->queue_head);
return 0;
}
return -EINVAL;
}
static struct block_device_operations nbd_fops =
{
.owner = THIS_MODULE,
.ioctl = nbd_ioctl,
};
/*
* And here should be modules and kernel interface
* (Just smiley confuses emacs :-)
*/
static int __init nbd_init(void)
{
int err = -ENOMEM;
int i;
if (sizeof(struct nbd_request) != 28) {
printk(KERN_CRIT "nbd: sizeof nbd_request needs to be 28 in order to work!\n" );
return -EIO;
}
if (nbds_max > MAX_NBD) {
printk(KERN_CRIT "nbd: cannot allocate more than %u nbds; %u requested.\n", MAX_NBD,
nbds_max);
return -EINVAL;
}
for (i = 0; i < nbds_max; i++) {
struct gendisk *disk = alloc_disk(1);
if (!disk)
goto out;
nbd_dev[i].disk = disk;
/*
* The new linux 2.5 block layer implementation requires
* every gendisk to have its very own request_queue struct.
* These structs are big so we dynamically allocate them.
*/
disk->queue = blk_init_queue(do_nbd_request, &nbd_lock);
if (!disk->queue) {
put_disk(disk);
goto out;
}
}
if (register_blkdev(NBD_MAJOR, "nbd")) {
err = -EIO;
goto out;
}
printk(KERN_INFO "nbd: registered device at major %d\n", NBD_MAJOR);
dprintk(DBG_INIT, "nbd: debugflags=0x%x\n", debugflags);
devfs_mk_dir("nbd");
for (i = 0; i < nbds_max; i++) {
struct gendisk *disk = nbd_dev[i].disk;
nbd_dev[i].file = NULL;
nbd_dev[i].magic = LO_MAGIC;
nbd_dev[i].flags = 0;
spin_lock_init(&nbd_dev[i].queue_lock);
INIT_LIST_HEAD(&nbd_dev[i].queue_head);
init_MUTEX(&nbd_dev[i].tx_lock);
[PATCH] nbd: fix TX/RX race condition Janos Haar of First NetCenter Bt. reported numerous crashes involving the NBD driver. With his help, this was tracked down to bogus bio vectors which in turn was the result of a race condition between the receive/transmit routines in the NBD driver. The bug manifests itself like this: CPU0 CPU1 do_nbd_request add req to queuelist nbd_send_request send req head for each bio kmap send nbd_read_stat nbd_find_request nbd_end_request kunmap When CPU1 finishes nbd_end_request, the request and all its associated bio's are freed. So when CPU0 calls kunmap whose argument is derived from the last bio, it may crash. Under normal circumstances, the race occurs only on the last bio. However, if an error is encountered on the remote NBD server (such as an incorrect magic number in the request), or if there were a bug in the server, it is possible for the nbd_end_request to occur any time after the request's addition to the queuelist. The following patch fixes this problem by making sure that requests are not added to the queuelist until after they have been completed transmission. In order for the receiving side to be ready for responses involving requests still being transmitted, the patch introduces the concept of the active request. When a response matches the current active request, its processing is delayed until after the tranmission has come to a stop. This has been tested by Janos and it has been successful in curing this race condition. From: Herbert Xu <herbert@gondor.apana.org.au> Here is an updated patch which removes the active_req wait in nbd_clear_queue and the associated memory barrier. I've also clarified this in the comment. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cc: <djani22@dynamicweb.hu> Cc: Paul Clements <Paul.Clements@SteelEye.com> Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-06 09:09:47 +01:00
init_waitqueue_head(&nbd_dev[i].active_wq);
nbd_dev[i].blksize = 1024;
nbd_dev[i].bytesize = 0x7ffffc00ULL << 10; /* 2TB */
disk->major = NBD_MAJOR;
disk->first_minor = i;
disk->fops = &nbd_fops;
disk->private_data = &nbd_dev[i];
disk->flags |= GENHD_FL_SUPPRESS_PARTITION_INFO;
sprintf(disk->disk_name, "nbd%d", i);
sprintf(disk->devfs_name, "nbd/%d", i);
set_capacity(disk, 0x7ffffc00ULL << 1); /* 2 TB */
add_disk(disk);
}
return 0;
out:
while (i--) {
blk_cleanup_queue(nbd_dev[i].disk->queue);
put_disk(nbd_dev[i].disk);
}
return err;
}
static void __exit nbd_cleanup(void)
{
int i;
for (i = 0; i < nbds_max; i++) {
struct gendisk *disk = nbd_dev[i].disk;
nbd_dev[i].magic = 0;
if (disk) {
del_gendisk(disk);
blk_cleanup_queue(disk->queue);
put_disk(disk);
}
}
devfs_remove("nbd");
unregister_blkdev(NBD_MAJOR, "nbd");
printk(KERN_INFO "nbd: unregistered device at major %d\n", NBD_MAJOR);
}
module_init(nbd_init);
module_exit(nbd_cleanup);
MODULE_DESCRIPTION("Network Block Device");
MODULE_LICENSE("GPL");
module_param(nbds_max, int, 0444);
MODULE_PARM_DESC(nbds_max, "How many network block devices to initialize.");
#ifndef NDEBUG
module_param(debugflags, int, 0644);
MODULE_PARM_DESC(debugflags, "flags for controlling debug output");
#endif