Skip to contentSkip to conversation

: message scheme tidy-up, tx_qlen, documentation

PATCH In review Gerrit Renker 2 participants 2 messages Nov 17, 2006

lore
Share & tools
lore
#0 lore raw

cover : message scheme tidy-up, tx_qlen, documentation

This includes mostly Ian's tx_qlen patch, the promised tidy-up of the WARN/BUG message scheme, and some trivial things.

Patch 1: This is Ian's tx_qlen updated tx_qlen patch.
I have removed the #define again (and was the one who suggested it); it seemed not to be a good idea after all (looking at net/dccp/dccp.h).

Patch 2: Adds the missing simplifications which were dropped somewhere inbetween
trying with and without using enum:8 for CCID 3 states. With that in place, many conditions can be simplified, as is now done.

The remaining patches are very trivial and have not been sent as a single patch in order to make the changes more clear when just reading through these.

Patch 3: Implements a consistent set of WARN/BUG messages as per list discussion.

Patch 4: Migration towards DCCP_BUG()/DCCP_BUG_ON(). The benefit is that these macros
will also inform about which condition was violated, in addition to line/file number.

Patch 5: Migration towards DCCP_WARN_ON(); same as in patch 4. Again, the difference is that
now the warning conditions are printed out as well.

Patch 6: Adds a rate-limited DCCP_WARN() wrapper.
In addition, it adds warning messages about RTT reaching zero in CCID 3 (as per earlier this discussions; maybe this should be promoted to DCCP_CRIT()).

Patch 7: After being brain-damaged from doing so many repetitive things, I thought to add
to this by completing the structure documentation for the rx/tx sockets. Ian can you please scream should I have erred somewhere (had one eye one the spec, though).

After this patch, the entire code has been updated with regard to this scheme. What has been left is BUG_TRAP(), which I think should be left since adding a DCCP-specific wrapper around it would not make much sense.
NB: Most of the ``severe-warning'' macros do not need a terminating linebreak.

Mapping wrt http://www.erg.abdn.ac.uk/users/gerrit/dccp/patch-backlog/ :

Patch 1     -  09z_tx_queue_len_via_sysctl_McDonald.diff
Patch 2..7  -  10{a..g}_*.diff
#1 lore raw
Quoted 14 lines · Gerrit Renker

This includes mostly Ian's tx_qlen patch, the promised tidy-up of the WARN/BUG message scheme, and some trivial things.

Patch 1: This is Ian's tx_qlen updated tx_qlen patch.
I have removed the #define again (and was the one who suggested it); it seemed not to be a good idea after all (looking at net/dccp/dccp.h).

Patch 2: Adds the missing simplifications which were dropped somewhere inbetween
trying with and without using enum:8 for CCID 3 states. With that in place, many conditions can be simplified, as is now done.

Quoted 6 lines

The remaining patches are very trivial and have not been sent as a single patch in order to make the changes more clear when just reading through these.

Patch 3: Implements a consistent set of WARN/BUG messages as per list discussion.

I only got patches 1 thru 3, any problems?

Gerrit, the "cherry-picking phase" is mostly over, your csets so far have been of great quality, so could you please consider using git so that I could just do a git-pull?

- Arnaldo