[PATCH] imsm: FIX: spare cannot be added

When Manage.c adds spare, it calls write_init_super() and this function
is responsible for closing disk handle (Manage.c:812).
For imsm case this handle is reused for managing this disk and handle
(due to this it is not closed).
As handle was opened with flag O_EXCL, adding disk to md fails on writing
to new_dev (md cannot set lock on device).
To resolve situation close current handle and open new one without O_EXCL flag.

Signed-off-by: Anna Czarnowska
Signed-off-by: Adam Kwolek
---

super-intel.c | 9 +++++++++
1 files changed, 9 insertions(+), 0 deletions(-)

diff --git a/super-intel.c b/super-intel.c
index 4cbb070..4e1be5d 100644
--- a/super-intel.c
+++ b/super-intel.c
@@ -3714,6 +3714,7 @@ static int mgmt_disk(struct supertype *st)
struct intel_super *super = st->sb;
size_t len;
struct imsm_update_add_remove_disk *u;
+ struct dl *d;

if (!super->disk_mgmt_list)
return 0;
@@ -3729,6 +3730,14 @@ static int mgmt_disk(struct supertype *st)
u->type = update_add_remove_disk;
append_metadata_update(st, u, len);

+ for (d = super->disk_mgmt_list; d ; d = d->next) {
+ char buf[PATH_MAX];
+
+ close(d->fd);
+ sprintf(buf, "%d:%d", d->major, d->minor);
+ d->fd = dev_open(buf, O_RDWR | O_DIRECT);
+ }
+
return 0;
}


--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: [PATCH] imsm: FIX: spare cannot be added

On Fri, 14 Jan 2011 13:13:04 +0100 Adam Kwolek wrote:

> When Manage.c adds spare, it calls write_init_super() and this function
> is responsible for closing disk handle (Manage.c:812).
> For imsm case this handle is reused for managing this disk and handle
> (due to this it is not closed).
> As handle was opened with flag O_EXCL, adding disk to md fails on writing
> to new_dev (md cannot set lock on device).
> To resolve situation close current handle and open new one without O_EXCL flag.
>
> Signed-off-by: Anna Czarnowska
> Signed-off-by: Adam Kwolek
> ---

Thanks for the patch.
I think the real problem is that write_init_super show not be closing
the fds. That seemed to make sense once, but it really doesn't now.
So I have:
- made sure that free_super always closed the fds
- stopped write_init_super from ever closing fds
- added a few more calls to free_super.

In particular, the bug you found is now simply fixed by calling
free_super before calling sysfs_add_disk.

Thanks,
NeilBrown

>
> super-intel.c | 9 +++++++++
> 1 files changed, 9 insertions(+), 0 deletions(-)
>
> diff --git a/super-intel.c b/super-intel.c
> index 4cbb070..4e1be5d 100644
> --- a/super-intel.c
> +++ b/super-intel.c
> @@ -3714,6 +3714,7 @@ static int mgmt_disk(struct supertype *st)
> struct intel_super *super = st->sb;
> size_t len;
> struct imsm_update_add_remove_disk *u;
> + struct dl *d;
>
> if (!super->disk_mgmt_list)
> return 0;
> @@ -3729,6 +3730,14 @@ static int mgmt_disk(struct supertype *st)
> u->type = update_add_remove_disk;
> append_metadata_update(st, u, len);
>
> + for (d = super->disk_mgmt_list; d ; d = d->next) {
> + char buf[PATH_MAX];
> +
> + close(d->fd);
> + sprintf(buf, "%d:%d", d->major, d->minor);
> + d->fd = dev_open(buf, O_RDWR | O_DIRECT);
> + }
> +
> return 0;
> }
>

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html