The limit was 128. This changes it to 1024.
--
Paul
On Tue, 29 Jan 2008 16:30:58 -0500
Paul Clements <[email protected]> wrote:
> The limit was 128. This changes it to 1024.
>
Please include the "why" in changelogs as well as the "what". It's fairly
easy to guess why this needs to be increased, but it would be useful and
interesting to hear about the operational scenarios in which 128 was
insufficient.
>
> --- ./include/linux/nbd.h.TIMEOUT 2007-08-22 13:18:12.000000000 -0400
> +++ ./include/linux/nbd.h 2008-01-29 14:35:17.000000000 -0500
> @@ -35,7 +35,7 @@ enum {
> };
>
> #define nbd_cmd(req) ((req)->cmd[0])
> -#define MAX_NBD 128
> +#define MAX_NBD 1024
>
> /* userspace doesn't need the nbd_device structure */
> #ifdef __KERNEL__
>
But one still needs to specify the nbds_max module parameter.
Can't we just nuke MAX_NBD altogether?
--- a/drivers/block/nbd.c~a
+++ a/drivers/block/nbd.c
@@ -54,7 +54,7 @@ static unsigned int debugflags;
#endif /* NDEBUG */
static unsigned int nbds_max = 16;
-static struct nbd_device nbd_dev[MAX_NBD];
+static struct nbd_device *nbd_dev;
/*
* Use just one lock (or at most 1 per NIC). Two arguments for this:
@@ -651,12 +651,6 @@ static int __init nbd_init(void)
BUILD_BUG_ON(sizeof(struct nbd_request) != 28);
- 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)
@@ -715,6 +709,10 @@ out:
static void __exit nbd_cleanup(void)
{
int i;
+
+ nbd_dev = kcalloc(nbds_max, sizeof(*nbd_dev), GFP_KERNEL);
+ if (!nbd_dev)
+ return -ENOMEM;
for (i = 0; i < nbds_max; i++) {
struct gendisk *disk = nbd_dev[i].disk;
nbd_dev[i].magic = 0;
diff -puN include/linux/nbd.h~a include/linux/nbd.h
--- a/include/linux/nbd.h~a
+++ a/include/linux/nbd.h
@@ -35,7 +35,6 @@ enum {
};
#define nbd_cmd(req) ((req)->cmd[0])
-#define MAX_NBD 128
/* userspace doesn't need the nbd_device structure */
#ifdef __KERNEL__
_
much nicer?
Andrew Morton wrote:
<snip>
> much nicer?
OK, yes, I was going for quick and easy, but you've got a point...
We do need to move the kcalloc into nbd_init instead of nbd_cleanup,
though -- that works a little better. Patch is attached. Thanks for the
suggestion.
--
Paul
On Tue, 29 Jan 2008 21:04:05 -0500 Paul Clements <[email protected]> wrote:
> Andrew Morton wrote:
>
> <snip>
>
> > much nicer?
>
> OK, yes, I was going for quick and easy, but you've got a point...
>
> We do need to move the kcalloc into nbd_init instead of nbd_cleanup,
> though -- that works a little better. Patch is attached. Thanks for the
> suggestion.
>
> --
> Paul
>
>
> [nbd_max_nbd_killed.diff text/x-patch (1.2KB)]
>
> Signed-Off-By: Paul Clements <[email protected]>
>
> --- ./include/linux/nbd.h.TIMEOUT 2007-08-22 13:18:12.000000000 -0400
> +++ ./include/linux/nbd.h 2008-01-29 20:01:33.000000000 -0500
Could we have a complete changelog please? As I mentioned in the
earlier email?
Andrew Morton wrote:
> On Tue, 29 Jan 2008 21:04:05 -0500 Paul Clements <[email protected]> wrote:
>
>> Andrew Morton wrote:
>>
>> <snip>
>>
>>> much nicer?
>> OK, yes, I was going for quick and easy, but you've got a point...
>>
>> We do need to move the kcalloc into nbd_init instead of nbd_cleanup,
>> though -- that works a little better. Patch is attached. Thanks for the
>> suggestion.
>>
>> --
>> Paul
>>
>>
>> [nbd_max_nbd_killed.diff text/x-patch (1.2KB)]
>>
>> Signed-Off-By: Paul Clements <[email protected]>
>>
>> --- ./include/linux/nbd.h.TIMEOUT 2007-08-22 13:18:12.000000000 -0400
>> +++ ./include/linux/nbd.h 2008-01-29 20:01:33.000000000 -0500
>
> Could we have a complete changelog please? As I mentioned in the
> earlier email?
>