Skip to contentSkip to conversation

docs: accounting/psi: drop stale 500ms window minimum from trigger docs

PATCH In review Tao Cui 2 participants 3 messages Jul 23 – Jul 28 Documentation

lore
Share & tools
lore

Tao Cui

author
#0 lore raw

PATCH 1/1 docs: accounting/psi: drop stale 500ms window minimum from trigger docs

From: Tao Cui <[email protected]>

psi.rst says trigger windows range "from 500ms to 10s" with a 50ms min polling interval. That minimum was removed in commit 519fabc7aaba ("psi:
remove 500ms min window size limitation for triggers"): the polling-rate protection moved to gating trigger creation behind CAP_SYS_RESOURCE, which also stopped capping psi event latency.

Update the wording accordingly: windows up to 10s with no enforced minimum; trigger creation requires CAP_SYS_RESOURCE on cgroup pressure files, while unprivileged users on /proc/pressure/ are limited to 2s multiples.

---
 Documentation/accounting/psi.rst | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)
Documentation/accounting/psi.rst +8 -9
index d455db3e5808..aae57fb5f15c 100644
@@ -99,15 +99,14 @@ psi metric and deactivates upon exit from the stall state. While system is
 in the stall state psi signal growth is monitored at a rate of 10 times per
 tracking window.
 
-The kernel accepts window sizes ranging from 500ms to 10s, therefore min
-monitoring update interval is 50ms and max is 1s. Min limit is set to
-prevent overly frequent polling. Max limit is chosen as a high enough number
-after which monitors are most likely not needed and psi averages can be used
-instead.
-
-Unprivileged users can also create monitors, with the only limitation that the
-window size must be a multiple of 2s, in order to prevent excessive resource
-usage.
+The kernel accepts window sizes up to 10s. The 10s maximum is chosen as a
+high enough number after which monitors are most likely not needed and psi
+averages can be used instead.
+
+To prevent excessive resource usage, trigger creation is restricted to
+privileged users (CAP_SYS_RESOURCE) on cgroup pressure files; on the
+system-wide /proc/pressure/ files unprivileged users can create monitors
+as well, limited to window sizes that are a multiple of 2s.
 
 When activated, psi monitor stays active for at least the duration of one
 tracking window to avoid repeated activations/deactivations when system is
Signature · 1 line
2.43.0
#1 lore raw
Quoted 34 lines · Tao Cui

From: Tao Cui <[email protected]>

psi.rst says trigger windows range "from 500ms to 10s" with a 50ms min polling interval. That minimum was removed in commit 519fabc7aaba ("psi:
remove 500ms min window size limitation for triggers"): the polling-rate protection moved to gating trigger creation behind CAP_SYS_RESOURCE, which also stopped capping psi event latency.

Update the wording accordingly: windows up to 10s with no enforced minimum; trigger creation requires CAP_SYS_RESOURCE on cgroup pressure files, while unprivileged users on /proc/pressure/ are limited to 2s multiples.

---
 Documentation/accounting/psi.rst | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)
Documentation/accounting/psi.rst +0 -7
index d455db3e5808..aae57fb5f15c 100644
@@ -99,15 +99,14 @@ psi metric and deactivates upon exit from the stall state. While system is
 in the stall state psi signal growth is monitored at a rate of 10 times per
 tracking window.
 
-The kernel accepts window sizes ranging from 500ms to 10s, therefore min
-monitoring update interval is 50ms and max is 1s. Min limit is set to
-prevent overly frequent polling. Max limit is chosen as a high enough number
-after which monitors are most likely not needed and psi averages can be used
-instead.
-
-Unprivileged users can also create monitors, with the only limitation that the

Documentation/accounting/psi.rst psi metric and deactivates upon exit from the stall state. While system is · line 109

Quoted 10 lines
Documentation/accounting/psi.rst +8 -2
-window size must be a multiple of 2s, in order to prevent excessive resource
-usage.
+The kernel accepts window sizes up to 10s. The 10s maximum is chosen as a
+high enough number after which monitors are most likely not needed and psi
+averages can be used instead.
+
+To prevent excessive resource usage, trigger creation is restricted to
+privileged users (CAP_SYS_RESOURCE) on cgroup pressure files; on the
+system-wide /proc/pressure/ files unprivileged users can create monitors
+as well, limited to window sizes that are a multiple of 2s.

Hmm. From this wording it sounds like we have different trigger creation rules for system-wide vs cgroup pressure files, which I think is not true. AFAICT, for both types, we restrict unprivileged users to creating triggers with 2s window size multiples.

Quoted 6 lines
Documentation/accounting/psi.rst +0 -1
 When activated, psi monitor stays active for at least the duration of one
 tracking window to avoid repeated activations/deactivations when system is
--

2.43.0

author
#2 lore raw
Quoted 42 lines · Suren Baghdasaryan
Quoted 41 lines · Tao Cui

From: Tao Cui <[email protected]>

psi.rst says trigger windows range "from 500ms to 10s" with a 50ms min polling interval. That minimum was removed in commit 519fabc7aaba ("psi:
remove 500ms min window size limitation for triggers"): the polling-rate protection moved to gating trigger creation behind CAP_SYS_RESOURCE, which also stopped capping psi event latency.

Update the wording accordingly: windows up to 10s with no enforced minimum; trigger creation requires CAP_SYS_RESOURCE on cgroup pressure files, while unprivileged users on /proc/pressure/ are limited to 2s multiples.

---
 Documentation/accounting/psi.rst | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)
Documentation/accounting/psi.rst +5 -9
index d455db3e5808..aae57fb5f15c 100644
@@ -99,15 +99,14 @@ psi metric and deactivates upon exit from the stall state. While system is
 in the stall state psi signal growth is monitored at a rate of 10 times per
 tracking window.
 
-The kernel accepts window sizes ranging from 500ms to 10s, therefore min
-monitoring update interval is 50ms and max is 1s. Min limit is set to
-prevent overly frequent polling. Max limit is chosen as a high enough number
-after which monitors are most likely not needed and psi averages can be used
-instead.
-
-Unprivileged users can also create monitors, with the only limitation that the
-window size must be a multiple of 2s, in order to prevent excessive resource
-usage.
+The kernel accepts window sizes up to 10s. The 10s maximum is chosen as a
+high enough number after which monitors are most likely not needed and psi
+averages can be used instead.
+
+To prevent excessive resource usage, trigger creation is restricted to
Quoted 9 lines

Documentation/accounting/psi.rst psi metric and deactivates upon exit from the stall state. While system is · line 109

Quoted 3 lines
Documentation/accounting/psi.rst +3 -0
+privileged users (CAP_SYS_RESOURCE) on cgroup pressure files; on the
+system-wide /proc/pressure/ files unprivileged users can create monitors
+as well, limited to window sizes that are a multiple of 2s.

Hmm. From this wording it sounds like we have different trigger creation rules for system-wide vs cgroup pressure files, which I think is not true. AFAICT, for both types, we restrict unprivileged users to creating triggers with 2s window size multiples.

Sorry, I missed part of the history here. 8b39d20eceed reverted the cgroup-specific gating that 519fabc7aaba added, so the 2s-multiple rule applies the same way to both system-wide and cgroup files. I'll send a v2 that keeps the window-range fix but restores the original unified wording.

Thanks,
Tao

Quoted 6 lines
Quoted 6 lines
Documentation/accounting/psi.rst +0 -1
 When activated, psi monitor stays active for at least the duration of one
 tracking window to avoid repeated activations/deactivations when system is
--

2.43.0