Archive-Team
/
zfs
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases 103 Wiki Activity
zfs/config/kernel-percpu.m4

88 lines
2.2 KiB
Plaintext
Raw Permalink Normal View History

Use percpu_counter for obj_alloc counter of Linux-backed caches A previous commit enabled the tracking of object allocations in Linux-backed caches from the SPL layer for debuggability. The commit is: 9a170fc6fe54f1e852b6c39630fe5ef2bbd97c16 Unfortunately, it also introduced minor performance regressions that were highlighted by the ZFS perf test-suite. Within Delphix we found that the regression would be from -1%, all the way up to -8% for some workloads. This commit brings performance back up to par by creating a separate counter for those caches and making it a percpu in order to avoid lock-contention. The initial performance testing was done by myself, and the final round was conducted by @tonynguien who was also the one that discovered the regression and highlighted the culprit. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Serapheim Dimitropoulos <serapheim@delphix.com> Closes #10397
2020-06-27 01:06:50 +00:00
dnl #
dnl # 3.18 API change,
dnl # The function percpu_counter_init now must be passed a GFP mask.
dnl #
AC_DEFUN([ZFS_AC_KERNEL_SRC_PERCPU_COUNTER_INIT], [
ZFS_LINUX_TEST_SRC([percpu_counter_init_with_gfp], [
#include <linux/gfp.h>
#include <linux/percpu_counter.h>
],[
struct percpu_counter counter;
int error;
error = percpu_counter_init(&counter, 0, GFP_KERNEL);
])
])
AC_DEFUN([ZFS_AC_KERNEL_PERCPU_COUNTER_INIT], [
AC_MSG_CHECKING([whether percpu_counter_init() wants gfp_t])
ZFS_LINUX_TEST_RESULT([percpu_counter_init_with_gfp], [
AC_MSG_RESULT(yes)
AC_DEFINE(HAVE_PERCPU_COUNTER_INIT_WITH_GFP, 1,
[percpu_counter_init() wants gfp_t])
],[
AC_MSG_RESULT(no)
])
])
Introduce write-mostly sums wmsum counters are a reduced version of aggsum counters, optimized for write-mostly scenarios. They do not provide optimized read functions, but instead allow much cheaper add function. The primary usage is infrequently read statistic counters, not requiring exact precision. The Linux implementation is directly mapped into percpu_counter KPI. The FreeBSD implementation is directly mapped into counter(9) KPI. In user-space due to lack of better implementation mapped to aggsum. Unfortunately neither Linux percpu_counter nor FreeBSD counter(9) provide sufficient functionality to completelly replace aggsum, so it still remains to be used for several hot counters. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@iXsystems.com> Signed-off-by: Alexander Motin <mav@FreeBSD.org> Sponsored-By: iXsystems, Inc. Closes #12114
2021-05-27 20:27:29 +00:00
dnl #
dnl # 4.13 API change,
dnl # __percpu_counter_add() was renamed to percpu_counter_add_batch().
dnl #
AC_DEFUN([ZFS_AC_KERNEL_SRC_PERCPU_COUNTER_ADD_BATCH], [
ZFS_LINUX_TEST_SRC([percpu_counter_add_batch], [
#include <linux/percpu_counter.h>
],[
struct percpu_counter counter;
percpu_counter_add_batch(&counter, 1, 1);
])
])
AC_DEFUN([ZFS_AC_KERNEL_PERCPU_COUNTER_ADD_BATCH], [
AC_MSG_CHECKING([whether percpu_counter_add_batch() is defined])
ZFS_LINUX_TEST_RESULT([percpu_counter_add_batch], [
AC_MSG_RESULT(yes)
AC_DEFINE(HAVE_PERCPU_COUNTER_ADD_BATCH, 1,
[percpu_counter_add_batch() is defined])
],[
AC_MSG_RESULT(no)
])
])
Linux 5.10 compat: percpu_ref added data member Kernel commit 2b0d3d3e4fcfb brought in some changes to the struct percpu_ref structure that moves most of its fields into a member struct named "data" of type struct percpu_ref_data. This includes the "count" member which is updated by vdev_blkg_tryget(), so update this function to chase the API change, and detect it via configure. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Coleman Kane <ckane@colemankane.org> Closes #11085
2020-10-18 14:36:12 +00:00
dnl #
dnl # 5.10 API change,
dnl # The "count" was moved into ref->data, from ref
dnl #
AC_DEFUN([ZFS_AC_KERNEL_SRC_PERCPU_REF_COUNT_IN_DATA], [
ZFS_LINUX_TEST_SRC([percpu_ref_count_in_data], [
#include <linux/percpu-refcount.h>
],[
struct percpu_ref_data d;
atomic_long_set(&d.count, 1L);
])
])
AC_DEFUN([ZFS_AC_KERNEL_PERCPU_REF_COUNT_IN_DATA], [
AC_MSG_CHECKING([whether is inside percpu_ref.data])
ZFS_LINUX_TEST_RESULT([percpu_ref_count_in_data], [
AC_MSG_RESULT(yes)
AC_DEFINE(ZFS_PERCPU_REF_COUNT_IN_DATA, 1,
[count is located in percpu_ref.data])
],[
AC_MSG_RESULT(no)
])
])
Use percpu_counter for obj_alloc counter of Linux-backed caches A previous commit enabled the tracking of object allocations in Linux-backed caches from the SPL layer for debuggability. The commit is: 9a170fc6fe54f1e852b6c39630fe5ef2bbd97c16 Unfortunately, it also introduced minor performance regressions that were highlighted by the ZFS perf test-suite. Within Delphix we found that the regression would be from -1%, all the way up to -8% for some workloads. This commit brings performance back up to par by creating a separate counter for those caches and making it a percpu in order to avoid lock-contention. The initial performance testing was done by myself, and the final round was conducted by @tonynguien who was also the one that discovered the regression and highlighted the culprit. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Serapheim Dimitropoulos <serapheim@delphix.com> Closes #10397
2020-06-27 01:06:50 +00:00
AC_DEFUN([ZFS_AC_KERNEL_SRC_PERCPU], [
ZFS_AC_KERNEL_SRC_PERCPU_COUNTER_INIT
Introduce write-mostly sums wmsum counters are a reduced version of aggsum counters, optimized for write-mostly scenarios. They do not provide optimized read functions, but instead allow much cheaper add function. The primary usage is infrequently read statistic counters, not requiring exact precision. The Linux implementation is directly mapped into percpu_counter KPI. The FreeBSD implementation is directly mapped into counter(9) KPI. In user-space due to lack of better implementation mapped to aggsum. Unfortunately neither Linux percpu_counter nor FreeBSD counter(9) provide sufficient functionality to completelly replace aggsum, so it still remains to be used for several hot counters. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@iXsystems.com> Signed-off-by: Alexander Motin <mav@FreeBSD.org> Sponsored-By: iXsystems, Inc. Closes #12114
2021-05-27 20:27:29 +00:00
ZFS_AC_KERNEL_SRC_PERCPU_COUNTER_ADD_BATCH
Linux 5.10 compat: percpu_ref added data member Kernel commit 2b0d3d3e4fcfb brought in some changes to the struct percpu_ref structure that moves most of its fields into a member struct named "data" of type struct percpu_ref_data. This includes the "count" member which is updated by vdev_blkg_tryget(), so update this function to chase the API change, and detect it via configure. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Coleman Kane <ckane@colemankane.org> Closes #11085
2020-10-18 14:36:12 +00:00
ZFS_AC_KERNEL_SRC_PERCPU_REF_COUNT_IN_DATA
Use percpu_counter for obj_alloc counter of Linux-backed caches A previous commit enabled the tracking of object allocations in Linux-backed caches from the SPL layer for debuggability. The commit is: 9a170fc6fe54f1e852b6c39630fe5ef2bbd97c16 Unfortunately, it also introduced minor performance regressions that were highlighted by the ZFS perf test-suite. Within Delphix we found that the regression would be from -1%, all the way up to -8% for some workloads. This commit brings performance back up to par by creating a separate counter for those caches and making it a percpu in order to avoid lock-contention. The initial performance testing was done by myself, and the final round was conducted by @tonynguien who was also the one that discovered the regression and highlighted the culprit. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Serapheim Dimitropoulos <serapheim@delphix.com> Closes #10397
2020-06-27 01:06:50 +00:00
])
AC_DEFUN([ZFS_AC_KERNEL_PERCPU], [
ZFS_AC_KERNEL_PERCPU_COUNTER_INIT
Introduce write-mostly sums wmsum counters are a reduced version of aggsum counters, optimized for write-mostly scenarios. They do not provide optimized read functions, but instead allow much cheaper add function. The primary usage is infrequently read statistic counters, not requiring exact precision. The Linux implementation is directly mapped into percpu_counter KPI. The FreeBSD implementation is directly mapped into counter(9) KPI. In user-space due to lack of better implementation mapped to aggsum. Unfortunately neither Linux percpu_counter nor FreeBSD counter(9) provide sufficient functionality to completelly replace aggsum, so it still remains to be used for several hot counters. Reviewed-by: Paul Dagnelie <pcd@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Reviewed-by: Ryan Moeller <ryan@iXsystems.com> Signed-off-by: Alexander Motin <mav@FreeBSD.org> Sponsored-By: iXsystems, Inc. Closes #12114
2021-05-27 20:27:29 +00:00
ZFS_AC_KERNEL_PERCPU_COUNTER_ADD_BATCH
Linux 5.10 compat: percpu_ref added data member Kernel commit 2b0d3d3e4fcfb brought in some changes to the struct percpu_ref structure that moves most of its fields into a member struct named "data" of type struct percpu_ref_data. This includes the "count" member which is updated by vdev_blkg_tryget(), so update this function to chase the API change, and detect it via configure. Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Coleman Kane <ckane@colemankane.org> Closes #11085
2020-10-18 14:36:12 +00:00
ZFS_AC_KERNEL_PERCPU_REF_COUNT_IN_DATA
Use percpu_counter for obj_alloc counter of Linux-backed caches A previous commit enabled the tracking of object allocations in Linux-backed caches from the SPL layer for debuggability. The commit is: 9a170fc6fe54f1e852b6c39630fe5ef2bbd97c16 Unfortunately, it also introduced minor performance regressions that were highlighted by the ZFS perf test-suite. Within Delphix we found that the regression would be from -1%, all the way up to -8% for some workloads. This commit brings performance back up to par by creating a separate counter for those caches and making it a percpu in order to avoid lock-contention. The initial performance testing was done by myself, and the final round was conducted by @tonynguien who was also the one that discovered the regression and highlighted the culprit. Reviewed-by: Matt Ahrens <matt@delphix.com> Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov> Signed-off-by: Serapheim Dimitropoulos <serapheim@delphix.com> Closes #10397
2020-06-27 01:06:50 +00:00
])