Kernel API autotools checks are slow #9132
Assignees
Labels
Type: Building
Indicates an issue related to building binaries
Type: Performance
Performance improvement or performance problem
Comments
ryao
added
Type: Building
|
another thing that would help i think is re-ordering some of the checks. a lot of the checks appear to be oldest kernel first with newest kernels at the end. which means the list gets longer and longer over time and it spends a lot of time "testing foo() with 2 params ... no", "testing foo with 3 params ... no", "with 4 params .. no", "with 5 ... yes". if we started with 5 then for most people things would probably be a lot faster just because there would be a lot less checks. |
|
The time doubles by redundant calls of |
|
It would be a pretty major redesign, but if several files could be built at once it could be sped up significantly. Calling into the Kbuild system is expensive, and calling it once (or more!) for each test is what's killing it. Being able to just make a single Makefile with all (most?) tests with a single |
I think this is worth investigating since if it can be made it work we can hopefully avoid the need to make extensive (highly error prone) changes to the existing checks. |
behlendorf
added a commit
to behlendorf/zfs
that referenced
this issue
Sep 21, 2019
Reduce the time required for ./configure to perform the needed
KABI checks by allowing kbuild to compile multiple test cases in
parallel. This was accomplished by splitting each test's source
code from the logic handling whether that code could be compiled
or not.
By introducing this split it's possible to minimize the number of
times kbuild needs to be invoked. As importantly, it means all of
the tests can be built in parallel. This does require a little extra
care since we expect some tests to fail, so the --keep-going (-k)
option must be provided otherwise some tests may not get compiled.
Furthermore, since a failure during the kbuild modpost phase will
result in an early exit; the final linking phase is limited to tests
which passed the initial compilation and produced an object file.
Once everything has been built the configure script proceeds as
previously. The only significant difference is that it now merely
needs to test for the existence of a .ko file to determine the
result of a given test. This vastly speeds up the entire process.
New test cases should use ZFS_LINUX_TEST_SRC to declare their test
source code and ZFS_LINUX_TEST_RESULT to check the result. All of
the existing kernel-*.m4 files have been updated accordingly, see
config/kernel-current-time.m4 for a basic example. The legacy
ZFS_LINUX_TRY_COMPILE macro has been kept to handle special cases
but it's use is not encouraged.
master (secs) patched (secs)
------------- ----------------
autogen.sh 61 68
configure 137 24 (~17% of current run time)
make -j $(nproc) 44 44
make rpms 287 150
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue openzfs#9132
behlendorf
added a commit
to behlendorf/zfs
that referenced
this issue
Sep 26, 2019
Reduce the time required for ./configure to perform the needed
KABI checks by allowing kbuild to compile multiple test cases in
parallel. This was accomplished by splitting each test's source
code from the logic handling whether that code could be compiled
or not.
By introducing this split it's possible to minimize the number of
times kbuild needs to be invoked. As importantly, it means all of
the tests can be built in parallel. This does require a little extra
care since we expect some tests to fail, so the --keep-going (-k)
option must be provided otherwise some tests may not get compiled.
Furthermore, since a failure during the kbuild modpost phase will
result in an early exit; the final linking phase is limited to tests
which passed the initial compilation and produced an object file.
Once everything has been built the configure script proceeds as
previously. The only significant difference is that it now merely
needs to test for the existence of a .ko file to determine the
result of a given test. This vastly speeds up the entire process.
New test cases should use ZFS_LINUX_TEST_SRC to declare their test
source code and ZFS_LINUX_TEST_RESULT to check the result. All of
the existing kernel-*.m4 files have been updated accordingly, see
config/kernel-current-time.m4 for a basic example. The legacy
ZFS_LINUX_TRY_COMPILE macro has been kept to handle special cases
but it's use is not encouraged.
master (secs) patched (secs)
------------- ----------------
autogen.sh 61 68
configure 137 24 (~17% of current run time)
make -j $(nproc) 44 44
make rpms 287 150
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue openzfs#9132
jgallag88
pushed a commit
to jgallag88/zfs
that referenced
this issue
Dec 6, 2019
Reduce the time required for ./configure to perform the needed
KABI checks by allowing kbuild to compile multiple test cases in
parallel. This was accomplished by splitting each test's source
code from the logic handling whether that code could be compiled
or not.
By introducing this split it's possible to minimize the number of
times kbuild needs to be invoked. As importantly, it means all of
the tests can be built in parallel. This does require a little extra
care since we expect some tests to fail, so the --keep-going (-k)
option must be provided otherwise some tests may not get compiled.
Furthermore, since a failure during the kbuild modpost phase will
result in an early exit; the final linking phase is limited to tests
which passed the initial compilation and produced an object file.
Once everything has been built the configure script proceeds as
previously. The only significant difference is that it now merely
needs to test for the existence of a .ko file to determine the
result of a given test. This vastly speeds up the entire process.
New test cases should use ZFS_LINUX_TEST_SRC to declare their test
source code and ZFS_LINUX_TEST_RESULT to check the result. All of
the existing kernel-*.m4 files have been updated accordingly, see
config/kernel-current-time.m4 for a basic example. The legacy
ZFS_LINUX_TRY_COMPILE macro has been kept to handle special cases
but it's use is not encouraged.
master (secs) patched (secs)
------------- ----------------
autogen.sh 61 68
configure 137 24 (~17% of current run time)
make -j $(nproc) 44 44
make rpms 287 150
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes openzfs#8547
Closes openzfs#9132
Closes openzfs#9341
tonyhutter
pushed a commit
to tonyhutter/zfs
that referenced
this issue
Dec 26, 2019
Reduce the time required for ./configure to perform the needed
KABI checks by allowing kbuild to compile multiple test cases in
parallel. This was accomplished by splitting each test's source
code from the logic handling whether that code could be compiled
or not.
By introducing this split it's possible to minimize the number of
times kbuild needs to be invoked. As importantly, it means all of
the tests can be built in parallel. This does require a little extra
care since we expect some tests to fail, so the --keep-going (-k)
option must be provided otherwise some tests may not get compiled.
Furthermore, since a failure during the kbuild modpost phase will
result in an early exit; the final linking phase is limited to tests
which passed the initial compilation and produced an object file.
Once everything has been built the configure script proceeds as
previously. The only significant difference is that it now merely
needs to test for the existence of a .ko file to determine the
result of a given test. This vastly speeds up the entire process.
New test cases should use ZFS_LINUX_TEST_SRC to declare their test
source code and ZFS_LINUX_TEST_RESULT to check the result. All of
the existing kernel-*.m4 files have been updated accordingly, see
config/kernel-current-time.m4 for a basic example. The legacy
ZFS_LINUX_TRY_COMPILE macro has been kept to handle special cases
but it's use is not encouraged.
master (secs) patched (secs)
------------- ----------------
autogen.sh 61 68
configure 137 24 (~17% of current run time)
make -j $(nproc) 44 44
make rpms 287 150
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes openzfs#8547
Closes openzfs#9132
Closes openzfs#9341
Conflicts:
Makefile.am
config/kernel-fpu.m4
tonyhutter
pushed a commit
to tonyhutter/zfs
that referenced
this issue
Dec 27, 2019
Reduce the time required for ./configure to perform the needed
KABI checks by allowing kbuild to compile multiple test cases in
parallel. This was accomplished by splitting each test's source
code from the logic handling whether that code could be compiled
or not.
By introducing this split it's possible to minimize the number of
times kbuild needs to be invoked. As importantly, it means all of
the tests can be built in parallel. This does require a little extra
care since we expect some tests to fail, so the --keep-going (-k)
option must be provided otherwise some tests may not get compiled.
Furthermore, since a failure during the kbuild modpost phase will
result in an early exit; the final linking phase is limited to tests
which passed the initial compilation and produced an object file.
Once everything has been built the configure script proceeds as
previously. The only significant difference is that it now merely
needs to test for the existence of a .ko file to determine the
result of a given test. This vastly speeds up the entire process.
New test cases should use ZFS_LINUX_TEST_SRC to declare their test
source code and ZFS_LINUX_TEST_RESULT to check the result. All of
the existing kernel-*.m4 files have been updated accordingly, see
config/kernel-current-time.m4 for a basic example. The legacy
ZFS_LINUX_TRY_COMPILE macro has been kept to handle special cases
but it's use is not encouraged.
master (secs) patched (secs)
------------- ----------------
autogen.sh 61 68
configure 137 24 (~17% of current run time)
make -j $(nproc) 44 44
make rpms 287 150
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes openzfs#8547
Closes openzfs#9132
Closes openzfs#9341
Conflicts:
Makefile.am
config/kernel-fpu.m4
stevijo
pushed a commit
to stevijo/zfs
that referenced
this issue
Jan 12, 2020
Reduce the time required for ./configure to perform the needed
KABI checks by allowing kbuild to compile multiple test cases in
parallel. This was accomplished by splitting each test's source
code from the logic handling whether that code could be compiled
or not.
By introducing this split it's possible to minimize the number of
times kbuild needs to be invoked. As importantly, it means all of
the tests can be built in parallel. This does require a little extra
care since we expect some tests to fail, so the --keep-going (-k)
option must be provided otherwise some tests may not get compiled.
Furthermore, since a failure during the kbuild modpost phase will
result in an early exit; the final linking phase is limited to tests
which passed the initial compilation and produced an object file.
Once everything has been built the configure script proceeds as
previously. The only significant difference is that it now merely
needs to test for the existence of a .ko file to determine the
result of a given test. This vastly speeds up the entire process.
New test cases should use ZFS_LINUX_TEST_SRC to declare their test
source code and ZFS_LINUX_TEST_RESULT to check the result. All of
the existing kernel-*.m4 files have been updated accordingly, see
config/kernel-current-time.m4 for a basic example. The legacy
ZFS_LINUX_TRY_COMPILE macro has been kept to handle special cases
but it's use is not encouraged.
master (secs) patched (secs)
------------- ----------------
autogen.sh 61 68
configure 137 24 (~17% of current run time)
make -j $(nproc) 44 44
make rpms 287 150
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes openzfs#8547
Closes openzfs#9132
Closes openzfs#9341
(cherry picked from commit 608f874)
Signed-off-by: Thomas Lamprecht <t.lamprecht@proxmox.com>
tonyhutter
pushed a commit
that referenced
this issue
Jan 23, 2020
Reduce the time required for ./configure to perform the needed
KABI checks by allowing kbuild to compile multiple test cases in
parallel. This was accomplished by splitting each test's source
code from the logic handling whether that code could be compiled
or not.
By introducing this split it's possible to minimize the number of
times kbuild needs to be invoked. As importantly, it means all of
the tests can be built in parallel. This does require a little extra
care since we expect some tests to fail, so the --keep-going (-k)
option must be provided otherwise some tests may not get compiled.
Furthermore, since a failure during the kbuild modpost phase will
result in an early exit; the final linking phase is limited to tests
which passed the initial compilation and produced an object file.
Once everything has been built the configure script proceeds as
previously. The only significant difference is that it now merely
needs to test for the existence of a .ko file to determine the
result of a given test. This vastly speeds up the entire process.
New test cases should use ZFS_LINUX_TEST_SRC to declare their test
source code and ZFS_LINUX_TEST_RESULT to check the result. All of
the existing kernel-*.m4 files have been updated accordingly, see
config/kernel-current-time.m4 for a basic example. The legacy
ZFS_LINUX_TRY_COMPILE macro has been kept to handle special cases
but it's use is not encouraged.
master (secs) patched (secs)
------------- ----------------
autogen.sh 61 68
configure 137 24 (~17% of current run time)
make -j $(nproc) 44 44
make rpms 287 150
Reviewed-by: Tony Hutter <hutter2@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #8547
Closes #9132
Closes #9341
Conflicts:
Makefile.am
config/kernel-fpu.m4
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Our Kernel API autotools checks invoke the kernel’s build system to build a kernel module as part of each check. I noticed that the Nvidia kernel driver also uses autotools and has a similar number of checks, but is much faster. It appears that Nvidia is invoking the compiler directly to speed up the process and that works for them. I imagine that might not work quite so well for testing for GPL symbol exports, but that technique likely could be adapted for the majority of our autotools checks to make things faster.
Speeding up the build system has always been a low priority for me, but I was recently troubleshooting what I thought to be a kernel bug that turned out to be the result of failing hardware causing miscompilation. While trying to figure it out, I had to sit through around a dozen full kernel and initramfs rebuilds for an entire day. At least an hour of that time seems to have been spent on autotools and after that experience, the build system performance suddenly feels like a higher priority to me. It is not high enough for me to do something right away, but it is high enough to have my attention. I am posting it here in case anyone wants to try to tackle this before I do. I am also self assigning it as a note to myself.
The text was updated successfully, but these errors were encountered: