mbox series

[v2,0/5] Assertion and debugging helpers

Message ID cover.1744881159.git.dsterba@suse.com (mailing list archive)
Headers show
Series Assertion and debugging helpers | expand

Message

David Sterba April 17, 2025, 9:16 a.m. UTC
v2:
- add the functionality of VASSERT to ASSERT proper

Hi,

this is a RFC series. We need to improve debugging and logging helpers
so there's no ASSERT(0) or the convoluted
WARN_ON(IS_DEFINED(CONFIG_BTRFS_DEBUG)). This was mentioned in past
discussions so here's my proposal.

The series is only build tested, I'd like to hear some feedback if this
is going the right direction or if there are suggestions for fine
tuning.

1) Update ASSERT macro, so we can print additional information when
it triggers, namely printing the values of the assertion expression.
More details in the first patch, basic pattern is something like

    ASSERT(value > limit, "value=%llu limit=%llu", value, limit);

The second patch shows it's application in volumes.c, converting about
half where it's relevant. There are about 800 assertions in fs/btrfs/
and we don't need to convert them all. This can be done incrementally
and as needed.

The verbose version is another macro, although with some preprocessor
magic it should be possible to make ASSERT take variable number of
arguments. Does not seem worth though.

2) Wrap WARN_ON(IS_DEFINED(CONFIG_BTRFS_DEBUG)) to DEBUG_WARN() with
optional message with printk format. This is used to replace the
WARN_ON(...) above and also the ASSERT(0).

The ultimate goal for me is to get rid of all ASSERT(0), it's not used
consistently and looks like it's a note to the code author. There may be
several reasons for it's use and although I've converted almost all to
DEBUG_WARN it may miss the intentions.

In some cases it may be better to add proper error handling, print a
message or warn and exit with error. Possibly the are cases where the
code cannot continue, meaning it should be a BUG_ON but this is also
something we want to convert to proper error handling.

David Sterba (5):
  btrfs: enhance ASSERT() to take optional format string
  btrfs: example use of enhanced ASSERT() in volumes.c
  btrfs: add debug build only WARN
  btrfs: convert WARN_ON(IS_ENABLED(CONFIG_BTRFS_DEBUG)) to DEBUG_WARN
  btrfs: convert ASSERT(0) to DEBUG_WARN()

 fs/btrfs/backref.c         |  4 +--
 fs/btrfs/delayed-ref.c     |  2 +-
 fs/btrfs/dev-replace.c     |  2 +-
 fs/btrfs/disk-io.c         |  2 +-
 fs/btrfs/extent-tree.c     |  2 +-
 fs/btrfs/extent_io.c       |  4 +--
 fs/btrfs/extent_map.c      |  2 +-
 fs/btrfs/free-space-tree.c | 27 +++++++++--------
 fs/btrfs/inode.c           |  6 ++--
 fs/btrfs/messages.h        | 59 +++++++++++++++++++++++++++++++-----
 fs/btrfs/qgroup.c          |  6 ++--
 fs/btrfs/relocation.c      |  4 +--
 fs/btrfs/send.c            |  4 +--
 fs/btrfs/space-info.c      |  2 +-
 fs/btrfs/tree-checker.c    |  8 ++---
 fs/btrfs/volumes.c         | 62 ++++++++++++++++++++++++--------------
 fs/btrfs/zoned.c           |  2 +-
 17 files changed, 129 insertions(+), 69 deletions(-)

Comments

Josef Bacik April 17, 2025, 12:51 p.m. UTC | #1
On Thu, Apr 17, 2025 at 11:16:58AM +0200, David Sterba wrote:
> v2:
> - add the functionality of VASSERT to ASSERT proper
> 
> Hi,
> 
> this is a RFC series. We need to improve debugging and logging helpers
> so there's no ASSERT(0) or the convoluted
> WARN_ON(IS_DEFINED(CONFIG_BTRFS_DEBUG)). This was mentioned in past
> discussions so here's my proposal.
> 
> The series is only build tested, I'd like to hear some feedback if this
> is going the right direction or if there are suggestions for fine
> tuning.
> 
> 1) Update ASSERT macro, so we can print additional information when
> it triggers, namely printing the values of the assertion expression.
> More details in the first patch, basic pattern is something like
> 
>     ASSERT(value > limit, "value=%llu limit=%llu", value, limit);
> 
> The second patch shows it's application in volumes.c, converting about
> half where it's relevant. There are about 800 assertions in fs/btrfs/
> and we don't need to convert them all. This can be done incrementally
> and as needed.
> 
> The verbose version is another macro, although with some preprocessor
> magic it should be possible to make ASSERT take variable number of
> arguments. Does not seem worth though.
> 
> 2) Wrap WARN_ON(IS_DEFINED(CONFIG_BTRFS_DEBUG)) to DEBUG_WARN() with
> optional message with printk format. This is used to replace the
> WARN_ON(...) above and also the ASSERT(0).
> 
> The ultimate goal for me is to get rid of all ASSERT(0), it's not used
> consistently and looks like it's a note to the code author. There may be
> several reasons for it's use and although I've converted almost all to
> DEBUG_WARN it may miss the intentions.
> 
> In some cases it may be better to add proper error handling, print a
> message or warn and exit with error. Possibly the are cases where the
> code cannot continue, meaning it should be a BUG_ON but this is also
> something we want to convert to proper error handling.
> 

I love it

Reviewed-by: Josef Bacik <josef@toxicpanda.com>

Thanks,

Josef