summaryrefslogtreecommitdiff
path: root/Documentation
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2017-05-11 21:01:56 +0300
committerLinus Torvalds <torvalds@linux-foundation.org>2017-05-11 21:01:56 +0300
commit55a1ab56c7981ab84373f372a4df75e3ee1535a7 (patch)
tree6859873da7c15e3760d6fd992e53516f84feadff /Documentation
parent9786e34e0a6055dbd1b46e16dfa791ac2b3da289 (diff)
parented6565e734249ef021d5c13ba34c167eb4e42f62 (diff)
downloadlinux-55a1ab56c7981ab84373f372a4df75e3ee1535a7.tar.xz
Merge branch 'for-linus' of git://git.kernel.dk/linux-block
Pull block fixes from Jens Axboe: "A smaller collection of fixes that should go into -rc1. This contains: - A fix from Christoph, fixing a regression with the WRITE_SAME and partial completions. Caused a BUG() on ppc. - Fixup for __blk_mq_stop_hw_queues(), it should be static. From Colin. - Removal of dmesg error messages on elevator switching, when invoked from sysfs. From me. - Fix for blk-stat, using this_cpu_ptr() in a section only protected by rcu_read_lock(). This breaks when PREEMPT_RCU is enabled. From me. - Two fixes for BFQ from Paolo, one fixing a crash and one updating the documentation. - An error handling lightnvm memory leak, from Rakesh. - The previous blk-mq hot unplug lock reversal depends on the CPU hotplug rework that isn't in mainline yet. This caused a lockdep splat when people unplugged CPUs with blk-mq devices. From Wanpeng. - A regression fix for DIF/DIX on blk-mq. From Wen" * 'for-linus' of git://git.kernel.dk/linux-block: block: handle partial completions for special payload requests blk-mq: NVMe 512B/4K+T10 DIF/DIX format returns I/O error on dd with split op blk-stat: don't use this_cpu_ptr() in a preemptable section elevator: remove redundant warnings on IO scheduler switch block, bfq: stress that low_latency must be off to get max throughput block, bfq: use pointer entity->sched_data only if set nvme: lightnvm: fix memory leak blk-mq: make __blk_mq_stop_hw_queues static lightnvm: remove unused rq parameter of nvme_nvm_rqtocmd() to kill warning block/mq: fix potential deadlock during cpu hotplug
Diffstat (limited to 'Documentation')
-rw-r--r--Documentation/block/bfq-iosched.txt17
1 files changed, 16 insertions, 1 deletions
diff --git a/Documentation/block/bfq-iosched.txt b/Documentation/block/bfq-iosched.txt
index 1b87df6cd476..05e2822a80b3 100644
--- a/Documentation/block/bfq-iosched.txt
+++ b/Documentation/block/bfq-iosched.txt
@@ -11,6 +11,13 @@ controllers), BFQ's main features are:
groups (switching back to time distribution when needed to keep
throughput high).
+In its default configuration, BFQ privileges latency over
+throughput. So, when needed for achieving a lower latency, BFQ builds
+schedules that may lead to a lower throughput. If your main or only
+goal, for a given device, is to achieve the maximum-possible
+throughput at all times, then do switch off all low-latency heuristics
+for that device, by setting low_latency to 0. Full details in Section 3.
+
On average CPUs, the current version of BFQ can handle devices
performing at most ~30K IOPS; at most ~50 KIOPS on faster CPUs. As a
reference, 30-50 KIOPS correspond to very high bandwidths with
@@ -375,11 +382,19 @@ default, low latency mode is enabled. If enabled, interactive and soft
real-time applications are privileged and experience a lower latency,
as explained in more detail in the description of how BFQ works.
-DO NOT enable this mode if you need full control on bandwidth
+DISABLE this mode if you need full control on bandwidth
distribution. In fact, if it is enabled, then BFQ automatically
increases the bandwidth share of privileged applications, as the main
means to guarantee a lower latency to them.
+In addition, as already highlighted at the beginning of this document,
+DISABLE this mode if your only goal is to achieve a high throughput.
+In fact, privileging the I/O of some application over the rest may
+entail a lower throughput. To achieve the highest-possible throughput
+on a non-rotational device, setting slice_idle to 0 may be needed too
+(at the cost of giving up any strong guarantee on fairness and low
+latency).
+
timeout_sync
------------