I am unable to ensure that the invariants such as having something in write_q upon restart are not violated, ever. So, let's rearrange the code a little bit to make sure.
[Before sending I noticed that this patch included a piece of other, unrelated patch. However, this passed tests on its own (because cli_wr_set_poll never returns an error). So in the interests of not screwing up further, sending as-is.] Signed-off-by: Pete Zaitcev <[email protected]> --- server/object.c | 6 ++++++ server/server.c | 9 ++++----- 2 files changed, 10 insertions(+), 5 deletions(-) commit 75d01d61e9ca55d244120cbbd7f338f07f03e9b6 Author: Master <[email protected]> Date: Thu May 20 21:26:55 2010 -0600 Protect the invariants. diff --git a/server/object.c b/server/object.c index 40ef4c4..7f6341a 100644 --- a/server/object.c +++ b/server/object.c @@ -342,6 +342,12 @@ bool object_get(struct client *cli, bool want_body) goto start_write; } + if (!cli->in_len) { + applog(LOG_INFO, "zero-sized object"); + cli_in_end(cli); + goto start_write; + } + if (!object_read_bytes(cli)) { cli_in_end(cli); return cli_err(cli, err, false); diff --git a/server/server.c b/server/server.c index 07d0375..de9e6a8 100644 --- a/server/server.c +++ b/server/server.c @@ -529,14 +529,13 @@ do_write: cli_wr_completed(cli, rc, &more_work); - if (more_work) - goto restart; - /* if we emptied the queue, clear write notification */ if (list_empty(&cli->write_q)) { cli->writing = false; - if (!cli_wr_set_poll(cli, false)) - goto err_out; + cli_wr_set_poll(cli, false); + } else { + if (more_work) + goto restart; } return; -- To unsubscribe from this list: send the line "unsubscribe hail-devel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html
