Hi Oliver On Thu, Aug 15, 2024 at 7:39 PM Oliver Sang <oliver.sang@xxxxxxxxx> wrote: > > hi, Jeff, > > On Thu, Aug 15, 2024 at 01:19:06PM -0700, Jeff Xu wrote: > > Hi Oliver, > > > > On Thu, Aug 15, 2024 at 11:16 AM Jeff Xu <jeffxu@xxxxxxxxxxxx> wrote: > > > > > > On Wed, Aug 14, 2024 at 12:14 AM <jeffxu@xxxxxxxxxxxx> wrote: > > > > > > > > From: Jeff Xu <jeffxu@xxxxxxxxxxxx> > > > > > > > > mremap doesn't allow relocate, expand, shrink across VMA boundaries, > > > > refactor the code to check src address range before doing anything on > > > > the destination, i.e. destination won't be unmapped, if src address > > > > failed the boundaries check. > > > > > > > > This also allows us to remove can_modify_mm from mremap.c, since > > > > the src address must be single VMA, can_modify_vma is used. > > > > > > > > It is likely this will improve the performance on mremap, previously > > > > the code does sealing check using can_modify_mm for the src address range, > > > > and the new code removed the loop (used by can_modify_mm). > > > > > > > > In order to verify this patch doesn't regress on mremap, I added tests in > > > > mseal_test, the test patch can be applied before mremap refactor patch or > > > > checkin independently. > > > > > > > > Also this patch doesn't change mseal's existing schematic: if sealing fail, > > > > user can expect the src/dst address isn't updated. So this patch can be > > > > applied regardless if we decided to go with current out-of-loop approach > > > > or in-loop approach currently in discussion. > > > > > > > > Regarding the perf test report by stress-ng [1] title: > > > > 8be7258aad: stress-ng.pagemove.page_remaps_per_sec -4.4% regression > > > > > > > > The test is using below for testing: > > > > stress-ng --timeout 60 --times --verify --metrics --no-rand-seed --pagemove 64 > > > > > > > > I can't repro this using ChromeOS, the pagemove test shows large value > > > > of stddev and stderr, and can't reasonably refect the performance impact. > > > > > > > > For example: I write a c program [2] to run the above pagemove test 10 times > > > > and calculate the stddev, stderr, for 3 commits: > > > > > > > > 1> before mseal feature is added: > > > > Ops/sec: > > > > Mean : 3564.40 > > > > Std Dev : 2737.35 (76.80% of Mean) > > > > Std Err : 865.63 (24.29% of Mean) > > > > > > > > 2> after mseal feature is added: > > > > Ops/sec: > > > > Mean : 2703.84 > > > > Std Dev : 2085.13 (77.12% of Mean) > > > > Std Err : 659.38 (24.39% of Mean) > > > > > > > > 3> after current patch (mremap refactor) > > > > Ops/sec: > > > > Mean : 3603.67 > > > > Std Dev : 2422.22 (67.22% of Mean) > > > > Std Err : 765.97 (21.26% of Mean) > > > > > > > > The result shows 21%-24% stderr, this means whatever perf improvment/impact > > > > there might be won't be measured correctly by this test. > > > > > > > > This test machine has 32G memory, Intel(R) Celeron(R) 7305, 5 CPU. > > > > And I reboot the machine before each test, and take the first 10 runs with > > > > run_stress_ng 10 > > > > > > > > (I will run longer duration to see if test still shows large stdDev,StdErr) > > > > > > > I took more samples (100 run ), the stddev/stderr is smaller, however > > > still not at a range that can reasonably measure the perf improvement > > > here. > > > > > > The tests were taken using the same machine as (10 times run above) > > > and exact the same steps: i.e. change to certain kernel commit, reboot > > > test device, take the first test result. > > > > > > 1> Before mseal feature is added: > > > Statistics: > > > Ops/sec: > > > Mean : 1733.26 > > > Std Dev : 842.13 (48.59% of Mean) > > > Std Err : 84.21 (4.86% of Mean) > > > > > > 2> After mseal feature is added > > > Statistics: > > > Ops/sec: > > > Mean : 1701.53 > > > Std Dev : 1017.29 (59.79% of Mean) > > > Std Err : 101.73 (5.98% of Mean) > > > > > > 3> After mremap refactor (this patch) > > > Statistics: > > > Ops/sec: > > > Mean : 1097.04 > > > Std Dev : 860.67 (78.45% of Mean) > > > Std Err : 86.07 (7.85% of Mean) > > > > > > Summary: even when the stderr is down to 4%-%8 percentage range, the > > > stddev is still too big. > > > > > > Hence, there are other unknown, random variables that impact this test. > > > > > I could not repro the 4% degradation with my test machine > > (Chromebook), this can be entirely due to the specific test and this > > test machine. > > > > Do you think it is possible to do a few more tests ? This time I like > > to have a larger sample size (100 run) > > > > stress-ng --timeout 60 --times --verify --metrics --no-rand-seed --pagemove 64 > > > > Please run the test for each commit following the exact steps, e.g. > > reboot the machine, run the test, get the first 100 results for > > sample. Please don't select or drop any unstable report because then > > the data will be biased. If possible, please includes stddiv and > > stderr for the data (or raw data if not possible, and I will do > > post-processing) > > > > for 3 commits: > > -> this patch. > > what's the base of it? could I directly apply this patch upon the commit > what you said "after mseal feature" as below? > > > -> after mseal feature > > -> before mseal feature > > could you exlictly point to two commit-id? sure this patch 8be7258a: mseal: add mseal syscall ff388fe5c: mseal: wire up mseal syscall > > > > Thank you for your time and assistance in helping me on understanding > > this issue. > > due to resource constraint, please expect that we need several days to finish > this test request. No problem. Thanks for your help! -Jeff > > > > Best regards, > > -Jeff > > > > > -Jeff > > > > > > > [1] https://lore.kernel.org/lkml/202408041602.caa0372-oliver.sang@xxxxxxxxx/ > > > > [2] https://github.com/peaktocreek/mmperf/blob/main/run_stress_ng.c > > > > > > > > > > > > Jeff Xu (2): > > > > mseal:selftest mremap across VMA boundaries. > > > > mseal: refactor mremap to remove can_modify_mm > > > > > > > > mm/internal.h | 24 ++ > > > > mm/mremap.c | 77 +++---- > > > > mm/mseal.c | 17 -- > > > > tools/testing/selftests/mm/mseal_test.c | 293 +++++++++++++++++++++++- > > > > 4 files changed, 353 insertions(+), 58 deletions(-) > > > > > > > > -- > > > > 2.46.0.76.ge559c4bf1a-goog > > > >
