You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
openwrt/target/linux/generic/pending-4.19/620-net_sched-codel-do-not-...

87 lines
3.0 KiB
Diff

From: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Date: Mon, 21 Aug 2017 11:14:14 +0300
Subject: [PATCH] net_sched/codel: do not defer queue length update
When codel wants to drop last packet in ->dequeue() it cannot call
qdisc_tree_reduce_backlog() right away - it will notify parent qdisc
about zero qlen and HTB/HFSC will deactivate class. The same class will
be deactivated second time by caller of ->dequeue(). Currently codel and
fq_codel defer update. This triggers warning in HFSC when it's qlen != 0
but there is no active classes.
This patch update parent queue length immediately: just temporary increase
qlen around qdisc_tree_reduce_backlog() to prevent first class deactivation
if we have skb to return.
This might open another problem in HFSC - now operation peek could fail and
deactivate parent class.
Signed-off-by: Konstantin Khlebnikov <khlebnikov@yandex-team.ru>
Link: https://bugzilla.kernel.org/show_bug.cgi?id=109581
---
--- a/net/sched/sch_codel.c
+++ b/net/sched/sch_codel.c
@@ -95,11 +95,17 @@ static struct sk_buff *codel_qdisc_deque
&q->stats, qdisc_pkt_len, codel_get_enqueue_time,
drop_func, dequeue_func);
- /* We cant call qdisc_tree_reduce_backlog() if our qlen is 0,
- * or HTB crashes. Defer it for next round.
+ /* If our qlen is 0 qdisc_tree_reduce_backlog() will deactivate
+ * parent class, dequeue in parent qdisc will do the same if we
+ * return skb. Temporary increment qlen if we have skb.
*/
- if (q->stats.drop_count && sch->q.qlen) {
- qdisc_tree_reduce_backlog(sch, q->stats.drop_count, q->stats.drop_len);
+ if (q->stats.drop_count) {
+ if (skb)
+ sch->q.qlen++;
+ qdisc_tree_reduce_backlog(sch, q->stats.drop_count,
+ q->stats.drop_len);
+ if (skb)
+ sch->q.qlen--;
q->stats.drop_count = 0;
q->stats.drop_len = 0;
}
--- a/net/sched/sch_fq_codel.c
+++ b/net/sched/sch_fq_codel.c
kernel: Make the patches apply on top of 4.19 This makes the patches which were just copied in the previous commit apply on top of kernel 4.19. The patches in the backports-4.19 folder were checked if they are really in kernel 4.19 based on the title and only removed if they were found in the upstream kernel. The following additional patches form the pending folder went into upstream Linux 4.19: pending-4.19/171-usb-dwc2-Fix-inefficient-copy-of-unaligned-buffers.patch pending-4.19/190-2-5-e1000e-Fix-wrong-comment-related-to-link-detection.patch pending-4.19/478-mtd-spi-nor-Add-support-for-XM25QH64A-and-XM25QH128A.patch pending-4.19/479-mtd-spi-nor-add-eon-en25qh32.patch pending-4.19/950-tty-serial-exar-generalize-rs485-setup.patch pending-4.19/340-MIPS-mm-remove-mips_dma_mapping_error.patch Bigger changes were introduced to the m25p80 spi nor driver, as far as I saw it in the new code, it now has the functionality provided in this patch: pending-4.19/450-mtd-m25p80-allow-fallback-from-spi_flash_read-to-reg.patch Part of this patch went upstream independent of OpenWrt: hack-4.19/220-gc_sections.patch This patch was reworked to match the changes done upstream. The MIPS DMA API changed a lot, this patch was rewritten to match the new DMA handling: pending-4.19/341-MIPS-mm-remove-no-op-dma_map_ops-where-possible.patch I did bigger manual changes to the following patches and I am not 100% sure if they are all correct: pending-4.19/0931-w1-gpio-fix-problem-with-platfom-data-in-w1-gpio.patch pending-4.19/411-mtd-partial_eraseblock_write.patch pending-4.19/600-netfilter_conntrack_flush.patch pending-4.19/611-netfilter_match_bypass_default_table.patch pending-4.19/670-ipv6-allow-rejecting-with-source-address-failed-policy.patch hack-4.19/211-host_tools_portability.patch hack-4.19/221-module_exports.patch hack-4.19/321-powerpc_crtsavres_prereq.patch hack-4.19/902-debloat_proc.patch This is based on patchset from Marko Ratkaj <marko.ratkaj@sartura.hr> Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
6 years ago
@@ -317,6 +317,21 @@ begin:
flow->dropped += q->cstats.drop_count - prev_drop_count;
flow->dropped += q->cstats.ecn_mark - prev_ecn_mark;
+ /* If our qlen is 0 qdisc_tree_reduce_backlog() will deactivate
+ * parent class, dequeue in parent qdisc will do the same if we
+ * return skb. Temporary increment qlen if we have skb.
+ */
+ if (q->cstats.drop_count) {
+ if (skb)
+ sch->q.qlen++;
+ qdisc_tree_reduce_backlog(sch, q->cstats.drop_count,
+ q->cstats.drop_len);
+ if (skb)
+ sch->q.qlen--;
+ q->cstats.drop_count = 0;
+ q->cstats.drop_len = 0;
+ }
+
if (!skb) {
/* force a pass through old_flows to prevent starvation */
if ((head == &q->new_flows) && !list_empty(&q->old_flows))
kernel: Make the patches apply on top of 4.19 This makes the patches which were just copied in the previous commit apply on top of kernel 4.19. The patches in the backports-4.19 folder were checked if they are really in kernel 4.19 based on the title and only removed if they were found in the upstream kernel. The following additional patches form the pending folder went into upstream Linux 4.19: pending-4.19/171-usb-dwc2-Fix-inefficient-copy-of-unaligned-buffers.patch pending-4.19/190-2-5-e1000e-Fix-wrong-comment-related-to-link-detection.patch pending-4.19/478-mtd-spi-nor-Add-support-for-XM25QH64A-and-XM25QH128A.patch pending-4.19/479-mtd-spi-nor-add-eon-en25qh32.patch pending-4.19/950-tty-serial-exar-generalize-rs485-setup.patch pending-4.19/340-MIPS-mm-remove-mips_dma_mapping_error.patch Bigger changes were introduced to the m25p80 spi nor driver, as far as I saw it in the new code, it now has the functionality provided in this patch: pending-4.19/450-mtd-m25p80-allow-fallback-from-spi_flash_read-to-reg.patch Part of this patch went upstream independent of OpenWrt: hack-4.19/220-gc_sections.patch This patch was reworked to match the changes done upstream. The MIPS DMA API changed a lot, this patch was rewritten to match the new DMA handling: pending-4.19/341-MIPS-mm-remove-no-op-dma_map_ops-where-possible.patch I did bigger manual changes to the following patches and I am not 100% sure if they are all correct: pending-4.19/0931-w1-gpio-fix-problem-with-platfom-data-in-w1-gpio.patch pending-4.19/411-mtd-partial_eraseblock_write.patch pending-4.19/600-netfilter_conntrack_flush.patch pending-4.19/611-netfilter_match_bypass_default_table.patch pending-4.19/670-ipv6-allow-rejecting-with-source-address-failed-policy.patch hack-4.19/211-host_tools_portability.patch hack-4.19/221-module_exports.patch hack-4.19/321-powerpc_crtsavres_prereq.patch hack-4.19/902-debloat_proc.patch This is based on patchset from Marko Ratkaj <marko.ratkaj@sartura.hr> Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
6 years ago
@@ -327,15 +342,6 @@ begin:
}
qdisc_bstats_update(sch, skb);
flow->deficit -= qdisc_pkt_len(skb);
- /* We cant call qdisc_tree_reduce_backlog() if our qlen is 0,
- * or HTB crashes. Defer it for next round.
- */
- if (q->cstats.drop_count && sch->q.qlen) {
- qdisc_tree_reduce_backlog(sch, q->cstats.drop_count,
- q->cstats.drop_len);
- q->cstats.drop_count = 0;
- q->cstats.drop_len = 0;
- }
return skb;
}