Re: block/bsg.c

[Jens Axboe <jens.axboe@oracle.com>]

On Mon, Jul 16 2007, Andrew Morton wrote:
> 
> A belated review (I've never seen this before and there it is in mainline)
> 
> > static char bsg_version[] = "block layer sg (bsg) 0.4";
> 
> `const' would be better.  That moves it into a write-protected memory section.

Agree

> > #define list_entry_bc(entry)	list_entry((entry), struct bsg_command, list)
> 
> This makes the code easier to write but harder to read.  We should optimise
> for readers.  Please open-code this at callsites.
> 
> Or at least convert it into a (commented) (possibly inlined) C function.

list_entry_to_bc(), then? The main objective is to save on typing, and
(just as important) make sure we don't bump over the 80 chars per line.

> > /*
> >  * just for testing
> >  */
> > #define BSG_MAJOR	(240)
> 
> What's this doing in mainline?  240 is a "reserved for local use" major. 
> This will cause collisions.  This code should be using dynamic major
> assignment.

Yeah, that's a big error on my part. Will get that fixed up right away.

> > static DEFINE_MUTEX(bsg_mutex);
> > static int bsg_device_nr, bsg_minor_idx;
> > 
> > #define BSG_LIST_SIZE	(8)
> 
> afacit this isn't really the size of a list.  It has something to do with
> the number of minors which are attached to that illegitimate major?  A
> comment here would help.  Perhaps this name is poorly chosen.

Yeah, it's the size of the array of lists :-)

> > #define bsg_list_idx(minor)	((minor) & (BSG_LIST_SIZE - 1))
> 
> Please prefer to write code in C, not in cpp.

Agree, that will die.

> > static struct hlist_head bsg_device_list[BSG_LIST_SIZE];
> 
> That is an array, not a list.

It's an array of lists.

> > static struct class *bsg_class;
> > static LIST_HEAD(bsg_class_list);
> > 
> > static struct kmem_cache *bsg_cmd_cachep;
> 
> How many of these items do we expect to be simultaneously allocated?  If
> that number is small then a custom kmem_cache is probably not warranted.

For your average desktop application, only a few are likely to be in
flight at the same time. For higher end stuff, it could be thousands.

> > /*
> >  * our internal command type
> >  */
> > struct bsg_command {
> > 	struct bsg_device *bd;
> > 	struct list_head list;
> > 	struct request *rq;
> > 	struct bio *bio;
> > 	struct bio *bidi_bio;
> > 	int err;
> > 	struct sg_io_v4 hdr;
> > 	struct sg_io_v4 __user *uhdr;
> > 	char sense[SCSI_SENSE_BUFFERSIZE];
> > };
> 
> Comments here, please.

Noted.

> > static void bsg_free_command(struct bsg_command *bc)
> > {
> > 	struct bsg_device *bd = bc->bd;
> > 	unsigned long flags;
> > 
> > 	kmem_cache_free(bsg_cmd_cachep, bc);
> > 
> > 	spin_lock_irqsave(&bd->lock, flags);
> > 	bd->queued_cmds--;
> > 	spin_unlock_irqrestore(&bd->lock, flags);
> > 
> > 	wake_up(&bd->wq_free);
> > }
> > 
> > static struct bsg_command *bsg_alloc_command(struct bsg_device *bd)
> > {
> > 	struct bsg_command *bc = ERR_PTR(-EINVAL);
> > 
> > 	spin_lock_irq(&bd->lock);
> > 
> > 	if (bd->queued_cmds >= bd->max_queue)
> > 		goto out;
> > 
> > 	bd->queued_cmds++;
> > 	spin_unlock_irq(&bd->lock);
> > 
> > 	bc = kmem_cache_alloc(bsg_cmd_cachep, GFP_USER);
> 
> This should be GFP_KERNEL.

Fixed

> > 	if (unlikely(!bc)) {
> > 		spin_lock_irq(&bd->lock);
> > 		bd->queued_cmds--;
> > 		bc = ERR_PTR(-ENOMEM);
> > 		goto out;
> > 	}
> > 
> > 	memset(bc, 0, sizeof(*bc));
> 
> Use kmem_cache_zalloc() above, remove this.

Fixed

> > 	bc->bd = bd;
> > 	INIT_LIST_HEAD(&bc->list);
> > 	dprintk("%s: returning free cmd %p\n", bd->name, bc);
> > 	return bc;
> > out:
> > 	spin_unlock_irq(&bd->lock);
> > 	return bc;
> > }
> > 
> > static inline void
> > bsg_del_done_cmd(struct bsg_device *bd, struct bsg_command *bc)
> > {
> > 	bd->done_cmds--;
> > 	list_del(&bc->list);
> > }
> 
> This only has a single caller.  It would be clearer to move this code into
> that caller.
>
> > static inline void
> > bsg_add_done_cmd(struct bsg_device *bd, struct bsg_command *bc)
> > {
> > 	bd->done_cmds++;
> > 	list_add_tail(&bc->list, &bd->done_list);
> > 	wake_up(&bd->wq_done);
> > }
> 
> Ditto.  Once this has been moved into the caller, that caller can then use
> the neater list_move().

Sure, why not (to both).

> > static inline int bsg_io_schedule(struct bsg_device *bd, int state)
> 
> This is too large to inline.

Fixed

> > {
> > 	DEFINE_WAIT(wait);
> > 	int ret = 0;
> > 
> > 	spin_lock_irq(&bd->lock);
> > 
> > 	BUG_ON(bd->done_cmds > bd->queued_cmds);
> > 
> > 	/*
> > 	 * -ENOSPC or -ENODATA?  I'm going for -ENODATA, meaning "I have no
> > 	 * work to do", even though we return -ENOSPC after this same test
> > 	 * during bsg_write() -- there, it means our buffer can't have more
> > 	 * bsg_commands added to it, thus has no space left.
> > 	 */
> > 	if (bd->done_cmds == bd->queued_cmds) {
> > 		ret = -ENODATA;
> > 		goto unlock;
> > 	}
> > 
> > 	if (!test_bit(BSG_F_BLOCK, &bd->flags)) {
> > 		ret = -EAGAIN;
> > 		goto unlock;
> > 	}
> > 
> > 	prepare_to_wait(&bd->wq_done, &wait, state);
> > 	spin_unlock_irq(&bd->lock);
> > 	io_schedule();
> > 	finish_wait(&bd->wq_done, &wait);
> 
> No, io_schedule() _must_ be called in state TASK_UNINTERRUPTIBLE.  If it
> gets called in state TASK_INTERRUPTIBLE then all the accounting which it
> does becomes wrong.
> 
> Fortunately the sole caller of this function _does_ use
> TASK_UNINTERRUPTIBLE.  The `state' arg to this function should be removed.

Fixed

> > 	if ((state == TASK_INTERRUPTIBLE) && signal_pending(current))
> > 		ret = -ERESTARTSYS;
> 
> And this code should be deleted.

Fixed

> > 	return ret;
> > unlock:
> > 	spin_unlock_irq(&bd->lock);
> > 	return ret;
> > }
> > 
> > static int blk_fill_sgv4_hdr_rq(request_queue_t *q, struct request *rq,
> > 				struct sg_io_v4 *hdr, int has_write_perm)
> > {
> > 	memset(rq->cmd, 0, BLK_MAX_CDB); /* ATAPI hates garbage after CDB */
> > 
> > 	if (copy_from_user(rq->cmd, (void *)(unsigned long)hdr->request,
> > 			   hdr->request_len))
> > 		return -EFAULT;
> > 
> > 	if (hdr->subprotocol == BSG_SUB_PROTOCOL_SCSI_CMD) {
> > 		if (blk_verify_command(rq->cmd, has_write_perm))
> > 			return -EPERM;
> > 	} else if (!capable(CAP_SYS_RAWIO))
> > 		return -EPERM;
> 
> As a reader of this code I'm wondering "hm, why is
> BSG_SUB_PROTOCOL_SCSI_CMD unprivileged, while other modes require
> CAP_SYS_RAWIO"?.
>
> This design/policy decision maybe was discussed on a mailing list
> somewhere, or even perhaps in a changelog (although I can't find it).  But
> it is so important, and is so unobvious from a reading of the code that I'd
> suggest that it is worth some discussion right here, in a code comment.
>
It's not unprivileged, it goes through the blk_verify_command() check
list.

> > 	/*
> > 	 * fill in request structure
> > 	 */
> > 	rq->cmd_len = hdr->request_len;
> > 	rq->cmd_type = REQ_TYPE_BLOCK_PC;
> > 
> > 	rq->timeout = (hdr->timeout * HZ) / 1000;
> > 	if (!rq->timeout)
> > 		rq->timeout = q->sg_timeout;
> > 	if (!rq->timeout)
> > 		rq->timeout = BLK_DEFAULT_SG_TIMEOUT;
> > 
> > 	return 0;
> > }
> > 
> > /*
> >  * Check if sg_io_v4 from user is allowed and valid
> >  */
> > static int
> > bsg_validate_sgv4_hdr(request_queue_t *q, struct sg_io_v4 *hdr, int *rw)
> > {
> > 	int ret = 0;
> > 
> > 	if (hdr->guard != 'Q')
> > 		return -EINVAL;
> 
> hm, "Q".  What is the user interface to this new stuff?

'Q' is just the magic identifier, like 'S' for sg v4.

> What does the code in bsg.c _do_, anyway??  Ho hum.

It's a driver for transporting sg v4 commands.

> > 	if (hdr->request_len > BLK_MAX_CDB)
> > 		return -EINVAL;
> > 	if (hdr->dout_xfer_len > (q->max_sectors << 9) ||
> > 	    hdr->din_xfer_len > (q->max_sectors << 9))
> 
> Are we sure that nothing here can exceed 4GB now and in the future?

We are far away from that in a single command currently and probably
ever.

> > 		return -EIO;
> > 
> > 	switch (hdr->protocol) {
> > 	case BSG_PROTOCOL_SCSI:
> > 		switch (hdr->subprotocol) {
> > 		case BSG_SUB_PROTOCOL_SCSI_CMD:
> > 		case BSG_SUB_PROTOCOL_SCSI_TRANSPORT:
> > 			break;
> > 		default:
> > 			ret = -EINVAL;
> > 		}
> > 		break;
> > 	default:
> > 		ret = -EINVAL;
> > 	}
> > 
> > 	*rw = hdr->dout_xfer_len ? WRITE : READ;
> > 	return ret;
> > }
> > 
> > /*
> >  * map sg_io_v4 to a request.
> >  */
> > static struct request *
> > bsg_map_hdr(struct bsg_device *bd, struct sg_io_v4 *hdr)
> > {
> > 	request_queue_t *q = bd->queue;
> > 	struct request *rq, *next_rq = NULL;
> > 	int ret, rw = 0; /* shut up gcc */
> 
> The modern way of shutting up gcc is uninitialized_var().

OK, I'll do that.

> > 	unsigned int dxfer_len;
> > 	void *dxferp = NULL;
> > 
> > 	dprintk("map hdr %llx/%u %llx/%u\n", (unsigned long long) hdr->dout_xferp,
> > 		hdr->dout_xfer_len, (unsigned long long) hdr->din_xferp,
> > 		hdr->din_xfer_len);
> > 
> > 	ret = bsg_validate_sgv4_hdr(q, hdr, &rw);
> > 	if (ret)
> > 		return ERR_PTR(ret);
> > 
> > 	/*
> > 	 * map scatter-gather elements seperately and string them to request
> > 	 */
> > 	rq = blk_get_request(q, rw, GFP_KERNEL);
> > 	if (!rq)
> > 		return ERR_PTR(-ENOMEM);
> > 	ret = blk_fill_sgv4_hdr_rq(q, rq, hdr, test_bit(BSG_F_WRITE_PERM,
> > 						       &bd->flags));
> > 	if (ret)
> > 		goto out;
> > 
> > 	if (rw == WRITE && hdr->din_xfer_len) {
> > 		if (!test_bit(QUEUE_FLAG_BIDI, &q->queue_flags)) {
> > 			ret = -EOPNOTSUPP;
> > 			goto out;
> > 		}
> > 
> > 		next_rq = blk_get_request(q, READ, GFP_KERNEL);
> > 		if (!next_rq) {
> > 			ret = -ENOMEM;
> > 			goto out;
> > 		}
> > 		rq->next_rq = next_rq;
> > 
> > 		dxferp = (void*)(unsigned long)hdr->din_xferp;
> 
> So... sg_io_v4.din_xferp is a user virtual address?
> 
> And `struct sg_io_v4' has just become part of the kernel ABI?  Beware that
> there is a move afoot to require test code, manpages and even LTP testcases
> for new ABI extensions.  Is this interface documented anywhere?

The documentation is likely very scarce atm, if anything. The command
layout was discussed at the storage summit and on linux-scsi.

> > 		ret =  blk_rq_map_user(q, next_rq, dxferp, hdr->din_xfer_len);
> > 		if (ret)
> > 			goto out;
> > 	}
> > 
> > 	if (hdr->dout_xfer_len) {
> > 		dxfer_len = hdr->dout_xfer_len;
> > 		dxferp = (void*)(unsigned long)hdr->dout_xferp;
> > 	} else if (hdr->din_xfer_len) {
> > 		dxfer_len = hdr->din_xfer_len;
> > 		dxferp = (void*)(unsigned long)hdr->din_xferp;
> > 	} else
> > 		dxfer_len = 0;
> > 
> > 	if (dxfer_len) {
> > 		ret = blk_rq_map_user(q, rq, dxferp, dxfer_len);
> > 		if (ret)
> > 			goto out;
> > 	}
> > 	return rq;
> > out:
> > 	blk_put_request(rq);
> > 	if (next_rq) {
> > 		blk_rq_unmap_user(next_rq->bio);
> > 		blk_put_request(next_rq);
> > 	}
> > 	return ERR_PTR(ret);
> > }
> > 
> > ...
> > 
> > static inline struct bsg_command *bsg_next_done_cmd(struct bsg_device *bd)
> > {
> > 	struct bsg_command *bc = NULL;
> > 
> > 	spin_lock_irq(&bd->lock);
> > 	if (bd->done_cmds) {
> > 		bc = list_entry_bc(bd->done_list.next);
> > 		bsg_del_done_cmd(bd, bc);
> > 	}
> > 	spin_unlock_irq(&bd->lock);
> > 
> > 	return bc;
> > }
> 
> This is too large to inline.

Fixed

> > static int blk_complete_sgv4_hdr_rq(struct request *rq, struct sg_io_v4 *hdr,
> > 				    struct bio *bio, struct bio *bidi_bio)
> > {
> > 	int ret = 0;
> > 
> > 	dprintk("rq %p bio %p %u\n", rq, bio, rq->errors);
> > 	/*
> > 	 * fill in all the output members
> > 	 */
> > 	hdr->device_status = status_byte(rq->errors);
> > 	hdr->transport_status = host_byte(rq->errors);
> > 	hdr->driver_status = driver_byte(rq->errors);
> > 	hdr->info = 0;
> > 	if (hdr->device_status || hdr->transport_status || hdr->driver_status)
> > 		hdr->info |= SG_INFO_CHECK;
> > 	hdr->din_resid = rq->data_len;
> > 	hdr->response_len = 0;
> > 
> > 	if (rq->sense_len && hdr->response) {
> > 		int len = min((unsigned int) hdr->max_response_len,
> > 			      rq->sense_len);
> 
> Use min_t here

Fixed

> > 		ret = copy_to_user((void*)(unsigned long)hdr->response,
> > 				   rq->sense, len);
> > 		if (!ret)
> > 			hdr->response_len = len;
> > 		else
> > 			ret = -EFAULT;
> > 	}
> > 
> > 	if (rq->next_rq) {
> > 		blk_rq_unmap_user(bidi_bio);
> > 		blk_put_request(rq->next_rq);
> > 	}
> > 
> > 	blk_rq_unmap_user(bio);
> > 	blk_put_request(rq);
> > 
> > 	return ret;
> > }
> > 
> > ...
> >
> > static ssize_t
> > __bsg_read(char __user *buf, size_t count, struct bsg_device *bd,
> > 	   const struct iovec *iov, ssize_t *bytes_read)
> > {
> > 	struct bsg_command *bc;
> > 	int nr_commands, ret;
> > 
> > 	if (count % sizeof(struct sg_io_v4))
> > 		return -EINVAL;
> > 
> > 	ret = 0;
> > 	nr_commands = count / sizeof(struct sg_io_v4);
> > 	while (nr_commands) {
> > 		bc = bsg_get_done_cmd(bd);
> > 		if (IS_ERR(bc)) {
> > 			ret = PTR_ERR(bc);
> > 			break;
> > 		}
> > 
> > 		/*
> > 		 * this is the only case where we need to copy data back
> > 		 * after completing the request. so do that here,
> > 		 * bsg_complete_work() cannot do that for us
> > 		 */
> > 		ret = blk_complete_sgv4_hdr_rq(bc->rq, &bc->hdr, bc->bio,
> > 					       bc->bidi_bio);
> > 
> > 		if (copy_to_user(buf, (char *) &bc->hdr, sizeof(bc->hdr)))
> > 			ret = -EFAULT;
> 
> Unneeded cast.

Fixed

> > 		bsg_free_command(bc);
> > 
> > 		if (ret)
> > 			break;
> > 
> > 		buf += sizeof(struct sg_io_v4);
> > 		*bytes_read += sizeof(struct sg_io_v4);
> > 		nr_commands--;
> > 	}
> > 
> > 	return ret;
> > }
> 
> This function returns zero or a negative errno (as should have been
> explainined in its covering comment).  Hence its return type of ssize_t is
> misleading.  It should return `int'.  Which is in fact the type of the
> local variable which it returns.

Fixed

> > static inline void bsg_set_block(struct bsg_device *bd, struct file *file)
> > {
> > 	if (file->f_flags & O_NONBLOCK)
> > 		clear_bit(BSG_F_BLOCK, &bd->flags);
> > 	else
> > 		set_bit(BSG_F_BLOCK, &bd->flags);
> > }
> > 
> > static inline void bsg_set_write_perm(struct bsg_device *bd, struct file *file)
> > {
> > 	if (file->f_mode & FMODE_WRITE)
> > 		set_bit(BSG_F_WRITE_PERM, &bd->flags);
> > 	else
> > 		clear_bit(BSG_F_WRITE_PERM, &bd->flags);
> > }
> 
> Still wondering what all this code does.  It _appears_ that the chosen user
> interface is via some device-special file?  And that an O_NONBLOCK open of
> that file has some special (undocumented?) significance?

There's no special meaning. IIRC, it's to avoid passing the file around.

> > static inline int err_block_err(int ret)
> > {
> > 	if (ret && ret != -ENOSPC && ret != -ENODATA && ret != -EAGAIN)
> > 		return 1;
> > 
> > 	return 0;
> > }
> 
> What a strange function.  The name is fairly meaningless.  A little comment
> would help decrease the mystery.

It is crap, will fix that.

> > static ssize_t
> > bsg_read(struct file *file, char __user *buf, size_t count, loff_t *ppos)
> > {
> > 	struct bsg_device *bd = file->private_data;
> > 	int ret;
> > 	ssize_t bytes_read;
> > 
> > 	dprintk("%s: read %Zd bytes\n", bd->name, count);
> > 
> > 	bsg_set_block(bd, file);
> > 	bytes_read = 0;
> > 	ret = __bsg_read(buf, count, bd, NULL, &bytes_read);
> > 	*ppos = bytes_read;
> > 
> > 	if (!bytes_read || (bytes_read && err_block_err(ret)))
> > 		bytes_read = ret;
> > 	return bytes_read;
> > }
> >
> > static ssize_t __bsg_write(struct bsg_device *bd, const char __user *buf,
> > 			   size_t count, ssize_t *bytes_read)
> > {
> > 	struct bsg_command *bc;
> > 	struct request *rq;
> > 	int ret, nr_commands;
> > 
> > 	if (count % sizeof(struct sg_io_v4))
> > 		return -EINVAL;
> > 
> > 	nr_commands = count / sizeof(struct sg_io_v4);
> > 	rq = NULL;
> > 	bc = NULL;
> > 	ret = 0;
> > 	while (nr_commands) {
> > 		request_queue_t *q = bd->queue;
> > 
> > 		bc = bsg_alloc_command(bd);
> > 		if (IS_ERR(bc)) {
> > 			ret = PTR_ERR(bc);
> > 			bc = NULL;
> > 			break;
> > 		}
> > 
> > 		bc->uhdr = (struct sg_io_v4 __user *) buf;
> > 		if (copy_from_user(&bc->hdr, buf, sizeof(bc->hdr))) {
> > 			ret = -EFAULT;
> > 			break;
> > 		}
> > 
> > 		/*
> > 		 * get a request, fill in the blanks, and add to request queue
> > 		 */
> > 		rq = bsg_map_hdr(bd, &bc->hdr);
> > 		if (IS_ERR(rq)) {
> > 			ret = PTR_ERR(rq);
> > 			rq = NULL;
> > 			break;
> > 		}
> > 
> > 		bsg_add_command(bd, q, bc, rq);
> > 		bc = NULL;
> > 		rq = NULL;
> > 		nr_commands--;
> > 		buf += sizeof(struct sg_io_v4);
> > 		*bytes_read += sizeof(struct sg_io_v4);
> > 	}
> > 
> > 	if (bc)
> > 		bsg_free_command(bc);
> > 
> > 	return ret;
> > }
> 
> Return type should be `int'.

Fixed

> > static ssize_t
> > bsg_write(struct file *file, const char __user *buf, size_t count, loff_t *ppos)
> > {
> > 	struct bsg_device *bd = file->private_data;
> > 	ssize_t bytes_read;
> 
> This variable should be called bytes_written.

Indeed, probably a copy-paste from bsg_read().

> > 	int ret;
> > 
> > 	dprintk("%s: write %Zd bytes\n", bd->name, count);
> > 
> > 	bsg_set_block(bd, file);
> > 	bsg_set_write_perm(bd, file);
> > 
> > 	bytes_read = 0;
> > 	ret = __bsg_write(bd, buf, count, &bytes_read);
> > 	*ppos = bytes_read;
> > 
> > 	/*
> > 	 * return bytes written on non-fatal errors
> > 	 */
> > 	if (!bytes_read || (bytes_read && err_block_err(ret)))
> > 		bytes_read = ret;
> > 	dprintk("%s: returning %Zd\n", bd->name, bytes_read);
> > 	return bytes_read;
> > }
> > 
> > ...
> 
> > 
> > static struct bsg_device *bsg_add_device(struct inode *inode,
> > 					 struct request_queue *rq,
> > 					 struct file *file)
> > {
> > 	struct bsg_device *bd = NULL;
> 
> Unneeded initialisation.

Fixed

> > #ifdef BSG_DEBUG
> > 	unsigned char buf[32];
> > #endif
> > 
> > 	bd = bsg_alloc_device();
> > 	if (!bd)
> > 		return ERR_PTR(-ENOMEM);
> > 
> > 	bd->queue = rq;
> > 	kobject_get(&rq->kobj);
> > 	bsg_set_block(bd, file);
> > 
> > 	atomic_set(&bd->ref_count, 1);
> > 	bd->minor = iminor(inode);
> > 	mutex_lock(&bsg_mutex);
> > 	hlist_add_head(&bd->dev_list, &bsg_device_list[bsg_list_idx(bd->minor)]);
> > 
> > 	strncpy(bd->name, rq->bsg_dev.class_dev->class_id, sizeof(bd->name) - 1);
> > 	dprintk("bound to <%s>, max queue %d\n",
> > 		format_dev_t(buf, inode->i_rdev), bd->max_queue);
> > 
> > 	mutex_unlock(&bsg_mutex);
> > 	return bd;
> > }
> > 
> > ...
> >
> > static int
> > bsg_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
> > 	  unsigned long arg)
> 
> This is an file_operations.ioctl() method.  It should instead have been
> implemented as an unlocked_ioctl handler.

Fixed.

> > {
> > 	struct bsg_device *bd = file->private_data;
> > 	int __user *uarg = (int __user *) arg;
> > 
> > 	if (!bd)
> > 		return -ENXIO;
> 
> This cannot happen, surely?

Nope, removed.

> > 	switch (cmd) {
> > 		/*
> > 		 * our own ioctls
> > 		 */
> > 	case SG_GET_COMMAND_Q:
> > 		return put_user(bd->max_queue, uarg);
> > 	case SG_SET_COMMAND_Q: {
> > 		int queue;
> > 
> > 		if (get_user(queue, uarg))
> > 			return -EFAULT;
> > 		if (queue < 1)
> > 			return -EINVAL;
> > 
> > 		spin_lock_irq(&bd->lock);
> > 		bd->max_queue = queue;
> > 		spin_unlock_irq(&bd->lock);
> > 		return 0;
> > 	}
> > 
> > 	/*
> > 	 * SCSI/sg ioctls
> > 	 */
> > 	case SG_GET_VERSION_NUM:
> > 	case SCSI_IOCTL_GET_IDLUN:
> > 	case SCSI_IOCTL_GET_BUS_NUMBER:
> > 	case SG_SET_TIMEOUT:
> > 	case SG_GET_TIMEOUT:
> > 	case SG_GET_RESERVED_SIZE:
> > 	case SG_SET_RESERVED_SIZE:
> > 	case SG_EMULATED_HOST:
> > 	case SCSI_IOCTL_SEND_COMMAND: {
> > 		void __user *uarg = (void __user *) arg;
> > 		return scsi_cmd_ioctl(file, bd->queue, NULL, cmd, uarg);
> > 	}
> > 	case SG_IO: {
> > 		struct request *rq;
> > 		struct bio *bio, *bidi_bio = NULL;
> > 		struct sg_io_v4 hdr;
> > 
> > 		if (copy_from_user(&hdr, uarg, sizeof(hdr)))
> > 			return -EFAULT;
> > 
> > 		rq = bsg_map_hdr(bd, &hdr);
> > 		if (IS_ERR(rq))
> > 			return PTR_ERR(rq);
> > 
> > 		bio = rq->bio;
> > 		if (rq->next_rq)
> > 			bidi_bio = rq->next_rq->bio;
> > 		blk_execute_rq(bd->queue, NULL, rq, 0);
> > 		blk_complete_sgv4_hdr_rq(rq, &hdr, bio, bidi_bio);
> > 
> > 		if (copy_to_user(uarg, &hdr, sizeof(hdr)))
> > 			return -EFAULT;
> > 
> > 		return 0;
> > 	}
> > 	/*
> > 	 * block device ioctls
> > 	 */
> > 	default:
> > #if 0
> > 		return ioctl_by_bdev(bd->bdev, cmd, arg);
> > #else
> > 		return -ENOTTY;
> > #endif
> > 	}
> > }
> 
> So we perform IO operations on this "device" by opening it and running
> ioctls, the args to which point to fairly complex data structures which lie
> in userspace and which contain addresses and lengths of userspace IO
> buffers?
> 
> What did Christoph think of this? :)

The sg v3 duplicates should go, it's remnant of when bsg was a sg v3
driver.

> > static struct file_operations bsg_fops = {
> > 	.read		=	bsg_read,
> > 	.write		=	bsg_write,
> > 	.poll		=	bsg_poll,
> > 	.open		=	bsg_open,
> > 	.release	=	bsg_release,
> > 	.ioctl		=	bsg_ioctl,
> 
> unlocked_ioctl, please.

Fixed

> > 	.owner		=	THIS_MODULE,
> > };
> > 
> > void bsg_unregister_queue(struct request_queue *q)
> > {
> > 	struct bsg_class_device *bcd = &q->bsg_dev;
> > 
> > 	if (!bcd->class_dev)
> > 		return;
> 
> Can this happen?

Probably only for double unregister, which means it should probably just
be a WARN_ON() or something instead.

> > 	mutex_lock(&bsg_mutex);
> > 	sysfs_remove_link(&q->kobj, "bsg");
> > 	class_device_destroy(bsg_class, MKDEV(BSG_MAJOR, bcd->minor));
> > 	bcd->class_dev = NULL;
> > 	list_del_init(&bcd->list);
> > 	bsg_device_nr--;
> > 	mutex_unlock(&bsg_mutex);
> > }
> > EXPORT_SYMBOL_GPL(bsg_unregister_queue);
> 
> <still wondering what all this code does>
> 
> Would I be correct in assuming that it offers services to device drivers,
> which have yet to be hooked up?

Yes. As mentioned many lines up, it is a SCSI generic type driver that
uses the (now) defined version 4 command structure. So it'll get hooked
up to ny capable device.

> > int bsg_register_queue(struct request_queue *q, const char *name)
> > {
> > 	struct bsg_class_device *bcd, *__bcd;
> > 	dev_t dev;
> > 	int ret = -EMFILE;
> > 	struct class_device *class_dev = NULL;
> > 
> > 	/*
> > 	 * we need a proper transport to send commands, not a stacked device
> > 	 */
> > 	if (!q->request_fn)
> > 		return 0;
> > 
> > 	bcd = &q->bsg_dev;
> > 	memset(bcd, 0, sizeof(*bcd));
> > 	INIT_LIST_HEAD(&bcd->list);
> > 
> > 	mutex_lock(&bsg_mutex);
> > 	if (bsg_device_nr == BSG_MAX_DEVS) {
> > 		printk(KERN_ERR "bsg: too many bsg devices\n");
> 
> 32768 is a lot of devices.  Why is there any limit at all?

It is pretty pointless killed.

> > 		goto err;
> > 	}
> > 
> > retry:
> > 	list_for_each_entry(__bcd, &bsg_class_list, list) {
> > 		if (__bcd->minor == bsg_minor_idx) {
> > 			bsg_minor_idx++;
> > 			if (bsg_minor_idx == BSG_MAX_DEVS)
> > 				bsg_minor_idx = 0;
> > 			goto retry;
> > 		}
> > 	}
> > 
> > 	bcd->minor = bsg_minor_idx++;
> > 	if (bsg_minor_idx == BSG_MAX_DEVS)
> > 		bsg_minor_idx = 0;
> 
> So what's happening here?  We're doing a linear, potentially O(n^2) search
> for a unique minor number?
> 
> I expect that you'll find that lib/idr.c provides a more elegant solution. 
> The tty code uses this, and there are other examples around the place.

idr will do nicely I think. I'll punt that to Tomo, he is the bsg
maintainer.

> > 	bcd->queue = q;
> > 	dev = MKDEV(BSG_MAJOR, bcd->minor);
> > 	class_dev = class_device_create(bsg_class, NULL, dev, bcd->dev, "%s", name);
> > 	if (IS_ERR(class_dev)) {
> > 		ret = PTR_ERR(class_dev);
> > 		goto err;
> > 	}
> > 	bcd->class_dev = class_dev;
> > 
> > 	if (q->kobj.sd) {
> > 		ret = sysfs_create_link(&q->kobj, &bcd->class_dev->kobj, "bsg");
> > 		if (ret)
> > 			goto err;
> > 	}
> > 
> > 	list_add_tail(&bcd->list, &bsg_class_list);
> > 	bsg_device_nr++;
> > 
> > 	mutex_unlock(&bsg_mutex);
> > 	return 0;
> > err:
> > 	if (class_dev)
> > 		class_device_destroy(bsg_class, MKDEV(BSG_MAJOR, bcd->minor));
> > 	mutex_unlock(&bsg_mutex);
> > 	return ret;
> > }
> > EXPORT_SYMBOL_GPL(bsg_register_queue);
> > 
> > ...
> >
> 
> 
> In terms of presentation: this code hit the tree as base patch plus what
> appear to be 20 bugfixes, none of which are really interesting or relevant
> to mainline.  Personally I think it would be nicer if all that out-of-tree
> development work was cleaned up and the new code goes in as a single hit.
> 
> This makes it a lot easier to find out "wtf does this code all do".  One
> finds the first commit and reads the changlog.  But this algorithm yields:
> 
>     bsg: support for full generic block layer SG v3
> 
> which is not helpful.

I agree, I did consider rebasing the merging all patches into a single
commit prior to submission. In retrospect that would have been better,
the bug fixes commits prior to inclusion is not that interesting.

-- 
Jens Axboe