| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAJZ5v0jaAN+UF6u8WLVXmO5k0MfFNL1ToM46bS5RbZ+Gwp3EWA@mail.gmail.com>
Date: Mon, 19 Jan 2026 13:09:08 +0100
From: "Rafael J. Wysocki" <rafael@...nel.org>
To: Christian Loehle <christian.loehle@....com>
Cc: "Rafael J. Wysocki" <rafael@...nel.org>, Linux PM <linux-pm@...r.kernel.org>,
LKML <linux-kernel@...r.kernel.org>, Doug Smythies <dsmythies@...us.net>
Subject: Re: [PATCH v1 0/5] cpuidle: governors: teo: Wakeup events
classification change and some refinements
On Mon, Jan 19, 2026 at 10:53 AM Christian Loehle
<christian.loehle@....com> wrote:
>
> On 1/16/26 12:29, Rafael J. Wysocki wrote:
> > On Fri, Jan 16, 2026 at 12:52 PM Christian Loehle
> > <christian.loehle@....com> wrote:
> >>
> >> On 1/14/26 19:42, Rafael J. Wysocki wrote:
> >>> Hi All,
> >>>
> >>> This material has been in my local queue for almost a full development cycle,
> >>> so time to post it.
> >>>
> >>> The motivation for the changes in this series is mostly theoretical, but I do
> >>> see some idle power improvements from patch [4/5], for example, but nothing
> >>> specifically worth reporting.
> >>>
> >>> The first patch simply prevents idle states with zero-size bins from being
> >>> selected sometimes when teo_select() runs with stopped tick.
> >>>
> >>> Patch [2/5] avoids counting tick wakeups as intercepts unless there are
> >>> sufficiently many intercepts within the tick period range to assume that
> >>> the tick wakeup may have clobbered a genuine intercept.
> >>>
> >>> Patch [3/5] simply updates a coefficient in one of the inequalities to be
> >>> somewhat easier to interpret (this should be a cosmetic change).
> >>>
> >>> Patch [4/5] changes the criteria used for classifying wakeup events as hits
> >>> or intercepts to (hopefully) make the classification work better for large
> >>> state bins.
> >>>
> >>> Patch [5/5] refines the idle state lookup based on intercepts to first
> >>> consider the state with the maximum intercepts metric, so that state is
> >>> always taken into consideration.
> >>>
> >>> Please see the individual patch changelogs for details.
> >>>
> >>> Thanks!
> >>>
> >>>
> >
> > Hi Christian,
> >
> >> Hi Rafael,
> >> I'll do the in-depth review, but have run some tests already.
> >> They are attached, platform is the usual rk3399.
> >> "teo" is mainline, "teo-$i" is with patches 1..$i applied.
> >
> > Thanks for testing!
> >
> >> There's a regression on teo-4 visible on the intercept heavy IO workloads,
> >> for idle misses that isn't strong enough to reflect in score changes except
> >> for the very slow mtdblock device.
> >> interestingly though there also seems to be a regression in
> >> mapper/dm-slow (dm device with 51ms delay on each IO), which is not
> >> intercept heavy.
> >> Looking at the state residencies it overuses the deepest state2 in
> >> where state1 was preferred for the other teo variants.
> >> I've attached that too for reference.
> >> I'm assuming that is because of the new intercept-logic-exclusion clause.
> >
> > So can you please restore that clause to its previous form, while
> > keeping the other changes in patch 4, and see if the regression is
> > still there?
> >
>
> Yep that restores it, I've attached the full results again with teo-4-p and teo-5-p
> being teo-4 and teo-5 but with
> - if (2 * idx_intercept_sum > cpu_data->total) {
> + if (2 * idx_intercept_sum > cpu_data->total - idx_hit_sum) {
>
> respectively.
>
> To highlight let me just quote the schbench results
> test gov i score %change idles idle_miss miss_rt belows aboves
> schbench teo-4 0 295.87 +5.51% 29014 1309 0.045 1168 141
> schbench teo-4 1 302.20 +7.76% 31642 1408 0.044 1105 303
> schbench teo-4 2 301.07 +7.36% 29966 1602 0.053 1393 209
> schbench teo-4 3 297.00 +5.91% 29982 1529 0.051 1266 263
> schbench teo-4 4 297.20 +5.98% 29432 1419 0.048 1194 225
> schbench teo-4 5 299.47 +6.79% 30262 1381 0.046 1189 192
> schbench teo-4 6 301.23 +7.42% 29810 1469 0.049 1234 235
> schbench teo-4 7 301.00 +7.34% 29916 1511 0.051 1355 156
> schbench teo-4 8 299.17 +6.68% 29246 1444 0.049 1296 148
> schbench teo-4 9 297.77 +6.18% 30306 1399 0.046 1195 204
> schbench teo-4 10 304.03 +8.42% 30904 1346 0.044 1163 183
> schbench teo-4 11 300.73 +7.24% 30716 1276 0.042 1089 187
> schbench teo-4 12 298.13 +6.31% 29872 1314 0.044 1081 233
> schbench teo-4 13 298.73 +6.53% 30386 1288 0.042 1092 196
> schbench teo-4 14 300.77 +7.25% 30150 1323 0.044 1074 249
> schbench teo-4 15 296.23 +5.63% 32178 1328 0.041 1061 267
> schbench teo-4 16 283.50 +1.09% 27472 1282 0.047 1126 156
> schbench teo-4 17 300.53 +7.17% 29644 1552 0.052 1375 177
> schbench teo-4 18 302.23 +7.77% 31456 1418 0.045 1194 224
> schbench teo-4 19 307.43 +9.63% 32518 1259 0.039 968 291
> schbench teo-4-p 0 277.53 -1.03% 28054 503 0.018 151 352
> schbench teo-4-p 1 304.07 +8.43% 30650 553 0.018 178 375
> schbench teo-4-p 2 273.73 -2.39% 27120 509 0.019 163 346
> schbench teo-4-p 3 300.20 +7.05% 29932 528 0.018 171 357
> schbench teo-4-p 4 300.00 +6.98% 29580 584 0.020 180 404
> schbench teo-4-p 5 296.03 +5.56% 30176 514 0.017 172 342
> schbench teo-4-p 6 300.50 +7.16% 29806 550 0.018 174 376
> schbench teo-4-p 7 271.07 -3.34% 26934 472 0.018 165 307
> schbench teo-4-p 8 299.00 +6.62% 30568 564 0.018 175 389
> schbench teo-4-p 9 300.80 +7.26% 29660 482 0.016 188 294
> schbench teo-4-p 10 300.57 +7.18% 30384 634 0.021 195 439
> schbench teo-4-p 11 296.53 +5.74% 29590 599 0.020 221 378
> schbench teo-4-p 12 298.77 +6.54% 30664 599 0.020 177 422
> schbench teo-4-p 13 302.47 +7.86% 30200 559 0.019 223 336
> schbench teo-4-p 14 300.43 +7.13% 30326 562 0.019 166 396
> schbench teo-4-p 15 292.87 +4.44% 28798 524 0.018 142 382
> schbench teo-4-p 16 299.63 +6.85% 29718 535 0.018 189 346
> schbench teo-4-p 17 302.10 +7.73% 31122 584 0.019 221 363
> schbench teo-4-p 18 294.13 +4.89% 28312 560 0.020 167 393
> schbench teo-4-p 19 302.20 +7.76% 29872 556 0.019 168 388
> schbench teo-5 0 299.53 +6.81% 31198 1312 0.042 1062 250
> schbench teo-5 1 296.67 +5.79% 30090 1383 0.046 1169 214
> schbench teo-5 2 299.60 +6.84% 30010 1566 0.052 1331 235
> schbench teo-5 3 300.00 +6.98% 30764 1466 0.048 1239 227
> schbench teo-5 4 298.77 +6.54% 30002 1383 0.046 1212 171
> schbench teo-5 5 299.30 +6.73% 29848 1481 0.050 1288 193
> schbench teo-5 6 298.90 +6.59% 29586 1441 0.049 1260 181
> schbench teo-5 7 300.37 +7.11% 30672 1460 0.048 1306 154
> schbench teo-5 8 301.00 +7.34% 31454 1383 0.044 1190 193
> schbench teo-5 9 298.80 +6.55% 30042 1415 0.047 1194 221
> schbench teo-5 10 302.23 +7.77% 30986 1305 0.042 1066 239
> schbench teo-5 11 297.30 +6.02% 30112 1405 0.047 1190 215
> schbench teo-5 12 299.33 +6.74% 29536 1482 0.050 1214 268
> schbench teo-5 13 288.87 +3.01% 28754 1244 0.043 1038 206
> schbench teo-5 14 300.17 +7.04% 30232 1517 0.050 1300 217
> schbench teo-5 15 301.43 +7.49% 30578 1576 0.052 1287 289
> schbench teo-5 16 282.10 +0.60% 29106 996 0.034 834 162
> schbench teo-5 17 278.03 -0.86% 27432 1123 0.041 926 197
> schbench teo-5 18 304.40 +8.55% 29934 1526 0.051 1296 230
> schbench teo-5 19 276.50 -1.40% 28072 1158 0.041 970 188
> schbench teo-5-p 0 299.47 +6.79% 29994 587 0.020 231 356
> schbench teo-5-p 1 298.27 +6.36% 31600 555 0.018 190 365
> schbench teo-5-p 2 297.20 +5.98% 30350 543 0.018 198 345
> schbench teo-5-p 3 294.30 +4.95% 29962 523 0.017 192 331
> schbench teo-5-p 4 303.30 +8.16% 30366 480 0.016 160 320
> schbench teo-5-p 5 298.87 +6.58% 29838 541 0.018 180 361
> schbench teo-5-p 6 301.63 +7.56% 30022 480 0.016 168 312
> schbench teo-5-p 7 298.47 +6.43% 30210 581 0.019 202 379
> schbench teo-5-p 8 298.77 +6.54% 30636 529 0.017 210 319
> schbench teo-5-p 9 298.63 +6.49% 30474 564 0.019 185 379
> schbench teo-5-p 10 300.10 +7.01% 30808 577 0.019 251 326
> schbench teo-5-p 11 295.87 +5.51% 30190 565 0.019 186 379
> schbench teo-5-p 12 301.27 +7.43% 29958 466 0.016 160 306
> schbench teo-5-p 13 280.83 +0.14% 27916 494 0.018 164 330
> schbench teo-5-p 14 294.23 +4.92% 30330 611 0.020 219 392
> schbench teo-5-p 15 297.13 +5.96% 30728 656 0.021 195 461
> schbench teo-5-p 16 297.97 +6.25% 30242 524 0.017 178 346
> schbench teo-5-p 17 301.27 +7.43% 31004 524 0.017 208 316
> schbench teo-5-p 18 297.97 +6.25% 29864 565 0.019 211 354
> schbench teo-5-p 19 298.43 +6.42% 30462 519 0.017 150 369
>
> The -p variants have less than half the idle_misses and fall into the same order of magnitude
> for this test than all the other teo variants.
Thanks for the data!
I'll send an update of patch 4 without the intercept-logic-exclusion clause.
I thought that it would be necessary, but that does not turn out to be the case.
Powered by blists - more mailing lists