git/list[1] front-page[2] threads[3] people[4] search[5] about
wed 2026-10-07 17:25 UTC

[PATCH v4 07/10] sideband: use writev(3p) to send pktlines

From
Patrick Steinhardt <ps@pks.im>
Date
Mar 13, 2026, 06:45 UTC
Message-ID
<20260313-pks-upload-pack-write-contention-v4-7-7a9668061f7f@pks.im>
In-Reply-To
<20260313-pks-upload-pack-write-contention-v4-0-7a9668061f7f@pks.im>

Every pktline that we send out via `send_sideband()` currently requires two syscalls: one to write the pktline's length, and one to send its data. This typically isn't all that much of a problem, but under extreme load the syscalls may cause contention in the kernel.

Refactor the code to instead use the newly introduced writev(3p) infra so that we can send out the data with a single syscall. This reduces the number of syscalls from around 133,000 calls to write(3p) to around 67,000 calls to writev(3p).

Suggested-by: Jeff King <peff@peff.net>
Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 sideband.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)
diff --git a/sideband.c b/sideband.c
index ea7c25211e..1ed6614eaf 100644
--- a/sideband.c
+++ b/sideband.c
@@ -264,6 +264,7 @@ void send_sideband(int fd, int band, const char *data, ssize_t sz, int packet_ma
 	const char *p = data;
 
 	while (sz) {
+		struct iovec iov[2];
 		unsigned n;
 		char hdr[5];
 
@@ -273,12 +274,19 @@ void send_sideband(int fd, int band, const char *data, ssize_t sz, int packet_ma
 		if (0 <= band) {
 			xsnprintf(hdr, sizeof(hdr), "%04x", n + 5);
 			hdr[4] = band;
-			write_or_die(fd, hdr, 5);
+			iov[0].iov_base = hdr;
+			iov[0].iov_len = 5;
 		} else {
 			xsnprintf(hdr, sizeof(hdr), "%04x", n + 4);
-			write_or_die(fd, hdr, 4);
+			iov[0].iov_base = hdr;
+			iov[0].iov_len = 4;
 		}
-		write_or_die(fd, p, n);
+
+		iov[1].iov_base = (void *) p;
+		iov[1].iov_len = n;
+
+		writev_or_die(fd, iov, ARRAY_SIZE(iov));
+
 		p += n;
 		sz -= n;
 	}
-- 
2.53.0.904.g2727be2e99.dirty
Previous: Patrick SteinhardtNext: Patrick Steinhardt
Message 8 of 11 in “upload-pack: reduce lock contention when writing packfile data”
  1. 00/10 upload-pack: reduce lock contention when writing packfile dataPatrick Steinhardt, Mar 13, 2026
  2. 01/10 upload-pack: fix debug statement when flushing packfile dataPatrick Steinhardt, Mar 13, 2026
  3. 02/10 upload-pack: adapt keepalives based on bufferingPatrick Steinhardt, Mar 13, 2026
  4. 03/10 upload-pack: prefer flushing data over sending keepalivePatrick Steinhardt, Mar 13, 2026
  5. 04/10 upload-pack: reduce lock contention when writing packfile dataPatrick Steinhardt, Mar 13, 2026
  6. 05/10 compat/posix: introduce writev(3p) wrapperPatrick Steinhardt, Mar 13, 2026
  7. 06/10 wrapper: introduce writev(3p) wrappersPatrick Steinhardt, Mar 13, 2026
  8. 07/10 sideband: use writev(3p) to send pktlinesPatrick Steinhardt, Mar 13, 2026
  9. 08/10 csum-file: introduce `hashfd_ext()`Patrick Steinhardt, Mar 13, 2026
  10. 09/10 csum-file: drop `hashfd_throughput()`Patrick Steinhardt, Mar 13, 2026
  11. 10/10 builtin/pack-objects: reduce lock contention when writing packfile dataPatrick Steinhardt, Mar 13, 2026

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.