Linus,
below is a patch for 2.4.0-test that adds documentation for
ll_rw_block and generic_make_request. It also corrects a type error
in a printk call.
I would really like to change a couple of names in the file:
"generic_make_request" would sound much better as "raw_rw_block".
that would highlight it's similarity to ll_rw_block, and would
remove the "generic" which seems to imply that it is possible to
replace it with a specific_make_request - which is wrong.
also
"__make_request" would sound much better as elevator_make_request, as
that is more descriptive of what it does. Also __foo() normally is
a special (e.g. no locking) case of foo(), but that is not the
case here.
Would you accept such as patch at this stage?
NeilBrown
--- drivers/block/ll_rw_blk.c 2000/12/03 23:59:07 1.1
+++ drivers/block/ll_rw_blk.c 2000/12/04 00:17:34
@@ -262,13 +262,18 @@
*
* Description:
* The normal way for &struct buffer_heads to be passed to a device driver
- * it to collect into requests on a request queue, and allow the device
- * driver to select requests off that queue when it is ready. This works
- * well for many block devices. However some block devices (typically
- * virtual devices such as md or lvm) do not benefit from the processes on
+ * is for them to be collected into requests on a request queue, and then to
+ * allow the device driver to select requests off that queue when it is ready.
+ * This works well for many block devices. However some block devices (typically
+ * virtual devices such as md or lvm) do not benefit from the processing on
* the request queue, and are served best by having the requests passed
* directly to them. This can be achieved by providing a function to
* blk_queue_make_request().
+ *
+ * Caveat:
+ * The driver that does this *must* be able to deal appropriately with
+ * buffers in "highmemory", either by calling bh_kmap() to get a kernel mapping,
+ * to by calling create_bounce() to create a buffer in normal memory.
**/
void blk_queue_make_request(request_queue_t * q, make_request_fn * mfn)
@@ -831,6 +836,40 @@
return 0;
}
+/**
+ * generic_make_request: hand a buffer head to it's device driver for I/O
+ * @rw: READ, WRITE, or READA - what sort of I/O is desired.
+ * @bh: The buffer head describing the location in memory and on the device.
+ *
+ * generic_make_request() is used to make I/O requests of block
+ * devices. It is passed a &struct buffer_head and a &rw value. The
+ * %READ and %WRITE options are (hopefully) obvious in meaning. The
+ * %READA value means that a read is required, but that the driver is
+ * free to fail the request if, for example, it cannot get needed
+ * resources immediately.
+ *
+ * generic_make_request() does not return any status. The
+ * success/failure status of the request, along with notification of
+ * completion, is delivered asynchronously through the bh->b_end_io
+ * function described (one day) else where.
+ *
+ * The caller of generic_make_request must make sure that b_page,
+ * b_addr, b_size are set to describe the memory buffer, that b_rdev
+ * and b_rsector are set to describe the device address, and the
+ * b_end_io and optionally b_private are set to describe how
+ * completion notification should be signaled. BH_Mapped should also
+ * be set (to confirm that b_dev and b_blocknr are valid).
+ *
+ * generic_make_request and the drivers it calls may use b_reqnext,
+ * and may change b_rdev and b_rsector. So the values of these fields
+ * should NOT be depended on after the call to generic_make_request.
+ * Because of this, the caller should record the device address
+ * information in b_dev and b_blocknr.
+ *
+ * Apart from those fields mentioned above, no other fields, and in
+ * particular, no other flags, are changed by generic_make_request or
+ * any lower level drivers.
+ * */
void generic_make_request (int rw, struct buffer_head * bh)
{
int major = MAJOR(bh->b_rdev);
@@ -851,7 +890,7 @@
when mounting a device. */
printk(KERN_INFO
"attempt to access beyond end of device\n");
- printk(KERN_INFO "%s: rw=%d, want=%d, limit=%d\n",
+ printk(KERN_INFO "%s: rw=%d, want=%ld, limit=%d\n",
kdevname(bh->b_rdev), rw,
(sector + count)>>1,
blk_size[major][MINOR(bh->b_rdev)]);
@@ -883,9 +922,39 @@
while (q->make_request_fn(q, rw, bh));
}
-/* This function can be used to request a number of buffers from a block
- device. Currently the only restriction is that all buffers must belong to
- the same device */
+/**
+ * ll_rw_block: low-level access to block devices
+ * @rw: whether to %READ or %WRITE or maybe %READA (readahead)
+ * @nr: number of &struct buffer_heads in the array
+ * @bhs: array of pointers to &struct buffer_head
+ *
+ * ll_rw_block() takes an array of pointers to &struct buffer_heads,
+ * and requests an I/O operation on them, either a %READ or a %WRITE.
+ * The third %READA option is described in the documentation for
+ * generic_make_request() which ll_rw_block() calls.
+ *
+ * This function provides extra functionality that is not in
+ * generic_make_request() that is relevant to buffers in the buffer
+ * cache or page cache. In particular it drops any buffer that it
+ * cannot get a lock on (with the BH_Lock state bit), any buffer that
+ * appears to be clean when doing a write request, and any buffer that
+ * appears to be up-to-date when doing read request. Further it marks
+ * as clean buffers that are processed for writing (the buffer cache
+ * wont assume that they are actually clean until the buffer gets
+ * unlocked).
+ *
+ * ll_rw_block sets the b_rdev and b_rsector for generic_make_request
+ * based on the values in b_dev and b_blocknr. It assumes that
+ * b_blocknr is in multiples of b_size bytes (b_rsector must always be
+ * in multiples of 512bytes).
+ *
+ * The comments in the documentation for generic_make_request() about
+ * what fields need to be set up apply to ll_rw_block as well.
+ *
+ * Caveat:
+ * All of the buffers must be for the same device, and must also be
+ * of the current approved size for the device.
+ */
void ll_rw_block(int rw, int nr, struct buffer_head * bhs[])
{
On Mon, 04 Dec 2000, Neil Brown wrote:
> Would you accept such as patch at this stage?
The patch seems to use too long a line length, is it possible to
reformat this with a right margin of 80 for the benefit of those who use
text-mode terminals to hack?
While I can't comment on the content, any documentation enhancements are
certainly welcome, thank you! :-)
> + * allow the device driver to select requests off that queue when it is ready.
> + * This works well for many block devices. However some block devices (typically
(Two samples of long lines).
--
Matthias Andree