Initial ramdisks do not work in linux-2.5.2{8,9}, because
fs/block_dev.c in these kernels has a new version of bd_open()
that does not set bdev->bd_inode->i_size when bdev->bd_openers is
non-zero.
I would appreciate information on whether this change in
bd_open's behavior is intended. If it is, then the following
patch makes updates the ramdisk driver to work again.
Also, I would appreciate knowing if anyone is acting as
maintainer for drivers/block/rd.c, or if I should just send
patches for rd.c directly to Linus if nobody complains on
the linux-kernel mailing list. I have some other minor patches for rd.c
to reduce its use of minor device numbers, and a patch that someone
whose name I don't remember posted long ago for dropping pages that
contiain all zeroes.
Adam J. Richter __ ______________ 575 Oroville Road
[email protected] \ / Milpitas, California 95035
+1 408 309-6081 | g g d r a s i l United States of America
"Free Software For The Rest Of Us."
--- linux-2.5.29/drivers/block/rd.c 2002-07-26 19:58:39.000000000 -0700
+++ linux/drivers/block/rd.c 2002-07-28 16:28:03.000000000 -0700
@@ -379,6 +404,7 @@
rd_bdev[unit]->bd_openers++;
rd_bdev[unit]->bd_block_size = rd_blocksize;
rd_bdev[unit]->bd_inode->i_mapping->a_ops = &ramdisk_aops;
+ rd_bdev[unit]->bd_inode->i_size = rd_kbsize[unit] << 10;
rd_bdev[unit]->bd_queue = &blk_dev[MAJOR_NR].request_queue;
}
On Sun, Jul 28, 2002 at 04:42:20PM -0700, Adam J. Richter wrote:
> Initial ramdisks do not work in linux-2.5.2{8,9}, because
Correct.
> Also, I would appreciate knowing if anyone is acting as
> maintainer for drivers/block/rd.c, or if I should just send
> patches for rd.c directly to Linus if nobody complains on
> the linux-kernel mailing list.
Al has been working on the problem. To permanently fix it so it doesn't
break each time a change to do_open() happens. Yes, this ramdisk driver
is that brittle.
> + rd_bdev[unit]->bd_inode->i_size = rd_kbsize[unit] << 10;
My temporary hack around the problem was to use rd_length[unit] since
that's already in bytes. But anything that adds code here just makes
the driver slightly more brittle.
--
Russell King ([email protected]) The developer of ARM Linux
http://www.arm.linux.org.uk/personal/aboutme.html
On Mon, 29 Jul 2002 00:51:56 +0100, Russell King wrote:
>On Sun, Jul 28, 2002 at 04:42:20PM -0700, Adam J. Richter wrote:
>> Initial ramdisks do not work in linux-2.5.2{8,9}, because
>Correct.
[...]
>Al has been working on the problem. To permanently fix it so it doesn't
>break each time a change to do_open() happens.
[...]
>My temporary hack around the problem was to use rd_length[unit] since
>that's already in bytes.
[...]
Thanks for the quick response. Your solution is simpler.
Although some bigger change may be in the works, would anyone
object to my submitting the following patch (your version of the fix)
to Linus in the meantime? I am cc'ing this message to Al Viro.
Adam J. Richter __ ______________ 575 Oroville Road
[email protected] \ / Milpitas, California 95035
+1 408 309-6081 | g g d r a s i l United States of America
"Free Software For The Rest Of Us."
--- linux-2.5.29/drivers/block/rd.c 2002-07-26 19:58:39.000000000 -0700
+++ linux/drivers/block/rd.c 2002-07-28 16:59:05.000000000 -0700
@@ -379,6 +404,7 @@
rd_bdev[unit]->bd_openers++;
rd_bdev[unit]->bd_block_size = rd_blocksize;
rd_bdev[unit]->bd_inode->i_mapping->a_ops = &ramdisk_aops;
+ rd_bdev[unit]->bd_inode->i_size = rd_length[unit];
rd_bdev[unit]->bd_queue = &blk_dev[MAJOR_NR].request_queue;
}