On Tue, Sep 11, 2018 at 04:11:45PM -0700, Omar Sandoval wrote:
> On Tue, Sep 11, 2018 at 01:57:39PM -0400, Josef Bacik wrote:
> > For enospc_debug having the block rsvs is super helpful to see if we've
> > done something wrong.
> >
> > Signed-off-by: Josef Bacik <jo...@toxicpanda.com>
> > ---
> > fs/btrfs/extent-tree.c | 16 ++++++++++++++++
> > 1 file changed, 16 insertions(+)
> >
> > diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
> > index a3baa16d456f..1cf66a92829b 100644
> > --- a/fs/btrfs/extent-tree.c
> > +++ b/fs/btrfs/extent-tree.c
> > @@ -7918,6 +7918,16 @@ static noinline int find_free_extent(struct
> > btrfs_fs_info *fs_info,
> > return ret;
> > }
> >
> > +static void dump_block_rsv(struct btrfs_fs_info *fs_info,
> > + struct btrfs_block_rsv *rsv)
> > +{
> > + spin_lock(&rsv->lock);
> > + btrfs_info(fs_info, "%d: size %llu reserved %llu\n",
no "\n" for the message helpers
> > + rsv->type, (unsigned long long)rsv->size,
> > + (unsigned long long)rsv->reserved);
u64 does not need the (unsigned long long) typecast since long.
> How about passing a string name for each of these instead of an ID which
> we have to cross-reference with the source?
Agreed, this would be better. As this is a debugging stuff, we can use
macros to simplify it like:
#define DUMP_BLOCK_RSV(fs_info, rsv_name) \
do { \
struct btrfs_block_rsv *__rsv = &(fs_info)->rsv_name; \
spin_lock(&__rsv->lock); \
btrfs_info((fs_info), #rsv_name ": size ..."); \
spin_unlock(&__rsv->lock); \
} while(0)
and use as
DUMP_BLOCK_RSV(fs_info, global_block_rsv);
The name of the block reserve name remains greppable.
