Skip to content

RFC relay revamp, third installment

Tom Zanussi lkml 1 version New

Get series with b4
b4 am 1222496261.6710.71.camel@charm-linux

Versions newest first

  1. v1 RFC Latest10 patches 0 repliesNew

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/10relay - Clean up relay_switch_subbuf() and make waking up consumers optional.include/linux/relay.h kernel/relay.c+75 -190 replies
2/10relay - Make the relay sub-buffer switch code replaceable.include/linux/relay.h kernel/relay.c+26 -100 replies
3/10relay - Add channel flags to relay, remove global callback param.block/blktrace.c include/linux/relay.h kernel/relay.c +1 more+27 -280 replies
4/10relay - Add reserved param to switch-subbuf, in preparation for non-pad write/reserve.include/linux/relay.h kernel/relay.c+25 -110 replies
5/10relay - Map the first sub-buffer at the end of the buffer, for temporary convenience.kernel/relay.c+15 -110 replies
6/10relay - Replace relay_reserve/relay_write with non-padded versions.include/linux/relay.h kernel/relay.c+64 -450 replies
7/10relay - Remove padding-related code from relay_read()/relay_splice_read() et al.kernel/relay.c+20 -1290 replies
8/10relay - Clean up remaining padding-related junk.block/blktrace.c include/linux/relay.h kernel/relay.c +1 more+13 -310 replies
9/10relay - Replace subbuf_start and notify_consumers callbacks with new_subbuf.block/blktrace.c include/linux/blktrace_api.h include/linux/relay.h +2 more+66 -950 replies
10/10relay - Remove sub-buffers from relay.block/blktrace.c include/linux/relay.h kernel/relay.c +1 more+15 -200 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.

Read the v1 thread

Cover letter [RFC PATCH 0/10] relay revamp, third installment

Here's the current relay cleanup patchset.

1-2 make the write path completely replaceable.
3 adds flags along with some related cleanup.
4-8 remove the padding in several stages.

The new patches in this set are:

9 simplifies the callbacks - now that we have flags, the subbuf_start callback is much simpler, has been combined with notify_consumers and has been renamed new_subbuf. Because part of the simplification has been to handle buffer-full conditions and count lost events internally, normal applications don't have to pay attention to it at all.

10 completely removes the idea of sub-buffers completely and now deals only with pages. relay_open() channges accordingly - buffer sizes are now in pages and consumers are woken only every n_wakeup pages, or never if this is 0.

It's a work in progress, but because I wanted the intermediate stages to actually work and not break anything, some of these patches, especially 05, are just temporary and will be removed in the next iteration.

I didn't have time to clean up the first 3 either - I'll also do that the next time around.

In the next round I plan to do vmap removal.

Tom