upstream/bind9
1
0
Fork 0
mirror of https://github.com/isc-projects/bind9.git synced 2026-05-28 04:34:54 -04:00
Code Issues Projects Releases Packages Wiki Activity Actions

Limit the number of DNS message processed from a single TCP read

Browse source 
The single TCP read can create as much as 64k divided by the minimum
size of the DNS message.  This can clog the processing thread and trash
the memory allocator because we need to do as much as ~20k allocations in
a single UV loop tick.

Limit the number of the DNS messages processed in a single UV loop tick
to just single DNS message and limit the number of the outstanding DNS
messages back to 23.  This effectively limits the number of pipelined
DNS messages to that number (this is the limit we already had before).

This reverts commit 780a89012d.
This commit is contained in:
Ondřej Surý 2024-06-05 09:15:39 +02:00 committed by Nicki Křížek
parent 0b1d70ed2a
commit 964891a794
No known key found for this signature in database
GPG key ID: 01623B9B652A20A7
2 changed files with 14 additions and 3 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

8
lib/isc/netmgr/netmgr-int.h
View file

@ -2262,6 +2262,14 @@ isc__nmsocket_readtimeout_cb(uv_timer_t *timer);
void
isc__nmsocket_writetimeout_cb(void *data, isc_result_t eresult);
/*%<
*
* Maximum number of simultaneous handles in flight supported for a single
* connected TCPDNS socket. This value was chosen arbitrarily, and may be
* changed in the future.
*/
#define STREAM_CLIENTS_PER_CONN 23
#define UV_RUNTIME_CHECK(func, ret) \
if (ret != 0) { \
FATAL_ERROR("%s failed: %s\n", #func, uv_strerror(ret)); \

9
lib/isc/netmgr/netmgr.c
View file

@ -2343,8 +2343,10 @@ processbuffer(isc_nmsocket_t *sock) {
* timers. If we do have a full message, reset the timer.
*
* Stop reading if this is a client socket, or if the server socket
* has been set to sequential mode. In this case we'll be called again
* later by isc__nm_resume_processing().
* has been set to sequential mode, or the number of queries we are
* processing simultaneously has reached the clients-per-connection
* limit. In this case we'll be called again later by
* isc__nm_resume_processing().
*/
isc_result_t
isc__nm_process_sock_buffer(isc_nmsocket_t *sock) {
@ -2384,7 +2386,8 @@ isc__nm_process_sock_buffer(isc_nmsocket_t *sock) {
isc__nmsocket_timer_stop(sock);
if (atomic_load(&sock->client) ||
atomic_load(&sock->sequential))
atomic_load(&sock->sequential) ||
atomic_load(&sock->ah) >= STREAM_CLIENTS_PER_CONN)
{
isc__nm_stop_reading(sock);
goto done;