CC: Jens Axboe <[email protected]>
Use the set_disk_ro() API when the backing file is read-only, to mark the disk
read-only, during the ->open(). The current hack does not work when doing a
mount -o remount.
Also, upgrade some warnings to WARN_ON(1) statements. They should actually
become BUG()s probably but I'll avoid that since I'm not so sure the change
works so well. I gave it only some limited testing.
Signed-off-by: Paolo 'Blaisorblade' Giarrusso <[email protected]>
---
linux-2.6.12-paolo/arch/um/drivers/ubd_kern.c | 14 ++++++++++----
1 files changed, 10 insertions(+), 4 deletions(-)
diff -puN arch/um/drivers/ubd_kern.c~uml-ubd-handle-readonly arch/um/drivers/ubd_kern.c
--- linux-2.6.12/arch/um/drivers/ubd_kern.c~uml-ubd-handle-readonly 2005-04-24 20:17:06.000000000 +0200
+++ linux-2.6.12-paolo/arch/um/drivers/ubd_kern.c 2005-04-24 20:17:06.000000000 +0200
@@ -156,6 +156,7 @@ static struct gendisk *fake_gendisk[MAX_
static struct openflags global_openflags = OPEN_FLAGS;
struct cow {
+ /* This is the backing file, actually */
char *file;
int fd;
unsigned long *bitmap;
@@ -927,10 +928,14 @@ static int ubd_open(struct inode *inode,
}
}
dev->count++;
- if((filp->f_mode & FMODE_WRITE) && !dev->openflags.w){
+ set_disk_ro(disk, !dev->openflags.w);
+
+ /* This should no more be needed. And it didn't work anyway to exclude
+ * read-write remounting of filesystems.*/
+ /*if((filp->f_mode & FMODE_WRITE) && !dev->openflags.w){
if(--dev->count == 0) ubd_close(dev);
err = -EROFS;
- }
+ }*/
out:
return(err);
}
@@ -1099,6 +1104,7 @@ static int prepare_request(struct reques
if((rq_data_dir(req) == WRITE) && !dev->openflags.w){
printk("Write attempted on readonly ubd device %s\n",
disk->disk_name);
+ WARN_ON(1); /* This should be impossible now */
end_request(req, 0);
return(1);
}
@@ -1252,6 +1258,7 @@ static int ubd_check_remapped(int fd, un
*/
printk("Write access to mapped page from readonly ubd "
"device %d\n", i);
+ WARN_ON(1); /* This should be impossible now */
return(0);
}
@@ -1605,8 +1612,7 @@ void do_io(struct io_thread_req *req)
}
} while((n < len) && (n != 0));
if (n < len) memset(&buf[n], 0, len - n);
- }
- else {
+ } else {
n = os_write_file(req->fds[bit], buf, len);
if(n != len){
printk("do_io - write failed err = %d "
_
On Sun, Apr 24 2005, [email protected] wrote:
> @@ -1099,6 +1104,7 @@ static int prepare_request(struct reques
> if((rq_data_dir(req) == WRITE) && !dev->openflags.w){
> printk("Write attempted on readonly ubd device %s\n",
> disk->disk_name);
> + WARN_ON(1); /* This should be impossible now */
> end_request(req, 0);
> return(1);
> }
I don't think that's a sound change. The WARN_ON() is strictly only
really useful for when you need the stack trace for something
interesting. As the io happens async, you will get a boring trace that
doesn't contain any valuable information.
--
Jens Axboe
On Monday 25 April 2005 12:16, Jens Axboe wrote:
> On Sun, Apr 24 2005, [email protected] wrote:
> > @@ -1099,6 +1104,7 @@ static int prepare_request(struct reques
> > if((rq_data_dir(req) == WRITE) && !dev->openflags.w){
> > printk("Write attempted on readonly ubd device %s\n",
> > disk->disk_name);
> > + WARN_ON(1); /* This should be impossible now */
> > end_request(req, 0);
> > return(1);
> > }
>
> I don't think that's a sound change. The WARN_ON() is strictly only
> really useful for when you need the stack trace for something
> interesting. As the io happens async, you will get a boring trace that
> doesn't contain any valuable information.
Ok, removed, and resending the patch, is the rest ok? I.e. is that supposed to
work? I gave a walk around and it seemed that the code handles
set_{disk,device}_ro() even during the open, but I'm no block layer expert.
Thanks for the review.
--
Paolo Giarrusso, aka Blaisorblade
Skype user "PaoloGiarrusso"
Linux registered user n. 292729
http://www.user-mode-linux.org/~blaisorblade
On Mon, Apr 25 2005, Blaisorblade wrote:
> On Monday 25 April 2005 12:16, Jens Axboe wrote:
> > On Sun, Apr 24 2005, [email protected] wrote:
> > > @@ -1099,6 +1104,7 @@ static int prepare_request(struct reques
> > > if((rq_data_dir(req) == WRITE) && !dev->openflags.w){
> > > printk("Write attempted on readonly ubd device %s\n",
> > > disk->disk_name);
> > > + WARN_ON(1); /* This should be impossible now */
> > > end_request(req, 0);
> > > return(1);
> > > }
> >
> > I don't think that's a sound change. The WARN_ON() is strictly only
> > really useful for when you need the stack trace for something
> > interesting. As the io happens async, you will get a boring trace that
> > doesn't contain any valuable information.
> Ok, removed, and resending the patch, is the rest ok? I.e. is that
> supposed to work? I gave a walk around and it seemed that the code
> handles set_{disk,device}_ro() even during the open, but I'm no block
> layer expert.
I'd keep the checks for sanity. Although the set_disk/device_ro prevents
regular fs write mounts, a buggy layered drive could still send down a
write by accident.
--
Jens Axboe