Dunfell is still at 2.11.2. Will there be a need for patches to this version too? Or should I take the lttng version bump along with the 5.4 version bumps that cause the issue?
Thanks! Steve On Mon, Sep 14, 2020 at 12:34 PM Bruce Ashfield <[email protected]> wrote: > > From: Jens Rehsack <[email protected]> > > Backporting the 10 patches since the lttng 2.12.2 release. We'll drop > them once .3 is released, but for now, we need the fixes to build > against the latest 5.4, 5.8 and 5.9 kernels. > > We also bump the devupstream SRCREV to pickup the same changes. > > Signed-off-by: Bruce Ashfield <[email protected]> > --- > ...ndency-issue-when-building-in-tree-w.patch | 54 ++ > ...ce.h-into-the-mmu-sub-directory-v5.9.patch | 41 + > ...Make-kvm_mmu_page-definition-and-acc.patch | 39 + > ...he-length-of-per-inode-prealloc-list.patch | 84 ++ > ...e-via-a-block-bitmap-read-is-prefetc.patch | 63 ++ > ...al-of-smp_-read_barrier_depends-v5.9.patch | 391 ++++++++ > ...teback-Drop-I_DIRTY_TIME_EXPIRE-v5.9.patch | 59 ++ > ...x-sync-livelock-due-to-b_dirty_time-.patch | 117 +++ > ...es-for-ext4_discard_preallocations-a.patch | 52 + > .../0010-Fix-system-call-filter-table.patch | 918 ++++++++++++++++++ > .../lttng/lttng-modules_2.12.2.bb | 12 +- > 11 files changed, 1829 insertions(+), 1 deletion(-) > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0001-Kconfig-fix-dependency-issue-when-building-in-tree-w.patch > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0002-fix-Move-mmutrace.h-into-the-mmu-sub-directory-v5.9.patch > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0003-fix-KVM-x86-mmu-Make-kvm_mmu_page-definition-and-acc.patch > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0004-fix-ext4-limit-the-length-of-per-inode-prealloc-list.patch > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0005-fix-ext4-indicate-via-a-block-bitmap-read-is-prefetc.patch > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0006-fix-removal-of-smp_-read_barrier_depends-v5.9.patch > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0007-fix-writeback-Drop-I_DIRTY_TIME_EXPIRE-v5.9.patch > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0008-fix-writeback-Fix-sync-livelock-due-to-b_dirty_time-.patch > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0009-fix-version-ranges-for-ext4_discard_preallocations-a.patch > create mode 100644 > meta/recipes-kernel/lttng/lttng-modules/0010-Fix-system-call-filter-table.patch > > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0001-Kconfig-fix-dependency-issue-when-building-in-tree-w.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0001-Kconfig-fix-dependency-issue-when-building-in-tree-w.patch > new file mode 100644 > index 0000000000..ae8bec45de > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0001-Kconfig-fix-dependency-issue-when-building-in-tree-w.patch > @@ -0,0 +1,54 @@ > +From ff4d1d7e85be94ef43709cd698f0ec9a12f247d1 Mon Sep 17 00:00:00 2001 > +From: Beniamin Sandu <[email protected]> > +Date: Thu, 13 Aug 2020 16:24:39 +0300 > +Subject: [PATCH 01/10] Kconfig: fix dependency issue when building in-tree > + without CONFIG_FTRACE > + > +When building in-tree, one could disable CONFIG_FTRACE from kernel > +config which will leave CONFIG_TRACEPOINTS selected by LTTNG modules, > +but generate a lot of linker errors like below because it leaves out > +other stuff, e.g.: > + > +trace.c:(.text+0xd86b): undefined reference to `trace_event_buffer_reserve' > +ld: trace.c:(.text+0xd8de): undefined reference to > `trace_event_buffer_commit' > +ld: trace.c:(.text+0xd926): undefined reference to `event_triggers_call' > +ld: trace.c:(.text+0xd942): undefined reference to > `trace_event_ignore_this_pid' > +ld: net/mac80211/trace.o: in function > `trace_event_raw_event_drv_tdls_cancel_channel_switch': > + > +It appears to be caused by the fact that TRACE_EVENT macros in the Linux > +kernel depend on the Ftrace ring buffer as soon as CONFIG_TRACEPOINTS is > +enabled. > + > +Steps to reproduce: > + > +- Get a clone of an upstream stable kernel and use scripts/built-in.sh on it > + > +- Configure a standard x86-64 build, enable built-in LTTNG but disable > + CONFIG_FTRACE from Kernel Hacking-->Tracers using menuconfig > + > +- Build will fail at linking stage > + > +Upstream-Status: Backport > + > +Signed-off-by: Beniamin Sandu <[email protected]> > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +--- > + Kconfig | 2 +- > + 1 file changed, 1 insertion(+), 1 deletion(-) > + > +diff --git a/Kconfig b/Kconfig > +index acdab73..10eccff 100644 > +--- a/Kconfig > ++++ b/Kconfig > +@@ -2,7 +2,7 @@ > + > + config LTTNG > + tristate "LTTng support" > +- select TRACEPOINTS > ++ select TRACING > + help > + LTTng is an open source tracing framework for Linux. > + > +-- > +2.19.1 > + > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0002-fix-Move-mmutrace.h-into-the-mmu-sub-directory-v5.9.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0002-fix-Move-mmutrace.h-into-the-mmu-sub-directory-v5.9.patch > new file mode 100644 > index 0000000000..fab673b854 > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0002-fix-Move-mmutrace.h-into-the-mmu-sub-directory-v5.9.patch > @@ -0,0 +1,41 @@ > +From e10ab43dd0e425df5bc0ac763447664ed075ba05 Mon Sep 17 00:00:00 2001 > +From: Michael Jeanson <[email protected]> > +Date: Mon, 10 Aug 2020 11:22:05 -0400 > +Subject: [PATCH 02/10] fix: Move mmutrace.h into the mmu/ sub-directory > (v5.9) > + > + commit 33e3042dac6bcc33b80835f7d7b502b1d74c457c > + Author: Sean Christopherson <[email protected]> > + Date: Mon Jun 22 13:20:29 2020 -0700 > + > + KVM: x86/mmu: Move mmu_audit.c and mmutrace.h into the mmu/ sub-directory > + > + Move mmu_audit.c and mmutrace.h under mmu/ where they belong. > + > +Upstream-Status: Backport > + > +Change-Id: I582525ccca34e1e3bd62870364108a7d3e9df2e4 > +Signed-off-by: Michael Jeanson <[email protected]> > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +--- > + probes/lttng-probe-kvm-x86-mmu.c | 4 ++++ > + 1 file changed, 4 insertions(+) > + > +diff --git a/probes/lttng-probe-kvm-x86-mmu.c > b/probes/lttng-probe-kvm-x86-mmu.c > +index 37384a2..5a7ef1e 100644 > +--- a/probes/lttng-probe-kvm-x86-mmu.c > ++++ b/probes/lttng-probe-kvm-x86-mmu.c > +@@ -24,7 +24,11 @@ > + */ > + #include <wrapper/tracepoint.h> > + > ++#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0)) > ++#include <../../arch/x86/kvm/mmu/mmutrace.h> > ++#else > + #include <../../arch/x86/kvm/mmutrace.h> > ++#endif > + > + #undef TRACE_INCLUDE_PATH > + #undef TRACE_INCLUDE_FILE > +-- > +2.19.1 > + > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0003-fix-KVM-x86-mmu-Make-kvm_mmu_page-definition-and-acc.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0003-fix-KVM-x86-mmu-Make-kvm_mmu_page-definition-and-acc.patch > new file mode 100644 > index 0000000000..524631cc72 > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0003-fix-KVM-x86-mmu-Make-kvm_mmu_page-definition-and-acc.patch > @@ -0,0 +1,39 @@ > +From f16315cc45c4c6b880de541bb092ca18a13952b7 Mon Sep 17 00:00:00 2001 > +From: Michael Jeanson <[email protected]> > +Date: Mon, 10 Aug 2020 11:36:03 -0400 > +Subject: [PATCH 03/10] fix: KVM: x86/mmu: Make kvm_mmu_page definition and > + accessor internal-only (v5.9) > + > + commit 985ab2780164698ec6e7d73fad523d50449261dd > + Author: Sean Christopherson <[email protected]> > + Date: Mon Jun 22 13:20:32 2020 -0700 > + > + KVM: x86/mmu: Make kvm_mmu_page definition and accessor internal-only > + > + Make 'struct kvm_mmu_page' MMU-only, nothing outside of the MMU should > + be poking into the gory details of shadow pages. > + > +Upstream-Status: Backport > + > +Change-Id: Ia5c1b9c49c2b00dad1d5b17c50c3dc730dafda20 > +Signed-off-by: Michael Jeanson <[email protected]> > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +--- > + probes/lttng-probe-kvm-x86-mmu.c | 1 + > + 1 file changed, 1 insertion(+) > + > +diff --git a/probes/lttng-probe-kvm-x86-mmu.c > b/probes/lttng-probe-kvm-x86-mmu.c > +index 5a7ef1e..8f98186 100644 > +--- a/probes/lttng-probe-kvm-x86-mmu.c > ++++ b/probes/lttng-probe-kvm-x86-mmu.c > +@@ -25,6 +25,7 @@ > + #include <wrapper/tracepoint.h> > + > + #if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0)) > ++#include <../../arch/x86/kvm/mmu/mmu_internal.h> > + #include <../../arch/x86/kvm/mmu/mmutrace.h> > + #else > + #include <../../arch/x86/kvm/mmutrace.h> > +-- > +2.19.1 > + > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0004-fix-ext4-limit-the-length-of-per-inode-prealloc-list.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0004-fix-ext4-limit-the-length-of-per-inode-prealloc-list.patch > new file mode 100644 > index 0000000000..e29c07252c > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0004-fix-ext4-limit-the-length-of-per-inode-prealloc-list.patch > @@ -0,0 +1,84 @@ > +From 8fe742807e65af29dac3fea568ff93cbc5dd9a56 Mon Sep 17 00:00:00 2001 > +From: Michael Jeanson <[email protected]> > +Date: Mon, 24 Aug 2020 15:26:04 -0400 > +Subject: [PATCH 04/10] fix: ext4: limit the length of per-inode prealloc list > + (v5.9) > +MIME-Version: 1.0 > +Content-Type: text/plain; charset=UTF-8 > +Content-Transfer-Encoding: 8bit > + > +See upstream commit: > + > + commit 27bc446e2def38db3244a6eb4bb1d6312936610a > + Author: brookxu <[email protected]> > + Date: Mon Aug 17 15:36:15 2020 +0800 > + > + ext4: limit the length of per-inode prealloc list > + > + In the scenario of writing sparse files, the per-inode prealloc list may > + be very long, resulting in high overhead for ext4_mb_use_preallocated(). > + To circumvent this problem, we limit the maximum length of per-inode > + prealloc list to 512 and allow users to modify it. > + > + After patching, we observed that the sys ratio of cpu has dropped, and > + the system throughput has increased significantly. We created a process > + to write the sparse file, and the running time of the process on the > + fixed kernel was significantly reduced, as follows: > + > + Running time on unfixed kernel： > + [root@TENCENT64 ~]# time taskset 0x01 ./sparse /data1/sparce.dat > + real 0m2.051s > + user 0m0.008s > + sys 0m2.026s > + > + Running time on fixed kernel： > + [root@TENCENT64 ~]# time taskset 0x01 ./sparse /data1/sparce.dat > + real 0m0.471s > + user 0m0.004s > + sys 0m0.395s > + > +Upstream-Status: Backport > + > +Signed-off-by: Michael Jeanson <[email protected]> > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +Change-Id: I5169cb24853d4da32e2862a6626f1f058689b053 > +--- > + instrumentation/events/lttng-module/ext4.h | 15 +++++++++++++++ > + 1 file changed, 15 insertions(+) > + > +diff --git a/instrumentation/events/lttng-module/ext4.h > b/instrumentation/events/lttng-module/ext4.h > +index 5f7ab28..72ad4c9 100644 > +--- a/instrumentation/events/lttng-module/ext4.h > ++++ b/instrumentation/events/lttng-module/ext4.h > +@@ -460,6 +460,20 @@ LTTNG_TRACEPOINT_EVENT(ext4_mb_release_group_pa, > + ) > + #endif > + > ++#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0)) > ++LTTNG_TRACEPOINT_EVENT(ext4_discard_preallocations, > ++ TP_PROTO(struct inode *inode, unsigned int len, unsigned int needed), > ++ > ++ TP_ARGS(inode, len, needed), > ++ > ++ TP_FIELDS( > ++ ctf_integer(dev_t, dev, inode->i_sb->s_dev) > ++ ctf_integer(ino_t, ino, inode->i_ino) > ++ ctf_integer(unsigned int, len, len) > ++ ctf_integer(unsigned int, needed, needed) > ++ ) > ++) > ++#else > + LTTNG_TRACEPOINT_EVENT(ext4_discard_preallocations, > + TP_PROTO(struct inode *inode), > + > +@@ -470,6 +484,7 @@ LTTNG_TRACEPOINT_EVENT(ext4_discard_preallocations, > + ctf_integer(ino_t, ino, inode->i_ino) > + ) > + ) > ++#endif > + > + LTTNG_TRACEPOINT_EVENT(ext4_mb_discard_preallocations, > + TP_PROTO(struct super_block *sb, int needed), > +-- > +2.19.1 > + > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0005-fix-ext4-indicate-via-a-block-bitmap-read-is-prefetc.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0005-fix-ext4-indicate-via-a-block-bitmap-read-is-prefetc.patch > new file mode 100644 > index 0000000000..f76e9698c8 > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0005-fix-ext4-indicate-via-a-block-bitmap-read-is-prefetc.patch > @@ -0,0 +1,63 @@ > +From 52563d02a9234215b62c5f519aa1b5d8589ccd0a Mon Sep 17 00:00:00 2001 > +From: Michael Jeanson <[email protected]> > +Date: Mon, 24 Aug 2020 15:37:50 -0400 > +Subject: [PATCH 05/10] =?UTF-8?q?fix:=20ext4:=20indicate=20via=20a=20block?= > + =?UTF-8?q?=20bitmap=20read=20is=20prefetched=E2=80=A6=20(v5.9)?= > +MIME-Version: 1.0 > +Content-Type: text/plain; charset=UTF-8 > +Content-Transfer-Encoding: 8bit > + > +See upstream commit: > + > + commit ab74c7b23f3770935016e3eb3ecdf1e42b73efaa > + Author: Theodore Ts'o <[email protected]> > + Date: Wed Jul 15 11:48:55 2020 -0400 > + > + ext4: indicate via a block bitmap read is prefetched via a tracepoint > + > + Modify the ext4_read_block_bitmap_load tracepoint so that it tells us > + whether a block bitmap is being prefetched. > + > +Upstream-Status: Backport > + > +Signed-off-by: Michael Jeanson <[email protected]> > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +Change-Id: I0e5e2c5b8004223d0928235c092449ee16a940e1 > +--- > + instrumentation/events/lttng-module/ext4.h | 14 ++++++++++++++ > + 1 file changed, 14 insertions(+) > + > +diff --git a/instrumentation/events/lttng-module/ext4.h > b/instrumentation/events/lttng-module/ext4.h > +index 72ad4c9..4476abb 100644 > +--- a/instrumentation/events/lttng-module/ext4.h > ++++ b/instrumentation/events/lttng-module/ext4.h > +@@ -893,12 +893,26 @@ LTTNG_TRACEPOINT_EVENT_INSTANCE(ext4__bitmap_load, > ext4_mb_buddy_bitmap_load, > + TP_ARGS(sb, group) > + ) > + > ++#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0)) > ++LTTNG_TRACEPOINT_EVENT(ext4_read_block_bitmap_load, > ++ TP_PROTO(struct super_block *sb, unsigned long group, bool prefetch), > ++ > ++ TP_ARGS(sb, group, prefetch), > ++ > ++ TP_FIELDS( > ++ ctf_integer(dev_t, dev, sb->s_dev) > ++ ctf_integer(__u32, group, group) > ++ ctf_integer(bool, prefetch, prefetch) > ++ ) > ++) > ++#else > + LTTNG_TRACEPOINT_EVENT_INSTANCE(ext4__bitmap_load, > ext4_read_block_bitmap_load, > + > + TP_PROTO(struct super_block *sb, unsigned long group), > + > + TP_ARGS(sb, group) > + ) > ++#endif > + > + LTTNG_TRACEPOINT_EVENT_INSTANCE(ext4__bitmap_load, ext4_load_inode_bitmap, > + > +-- > +2.19.1 > + > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0006-fix-removal-of-smp_-read_barrier_depends-v5.9.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0006-fix-removal-of-smp_-read_barrier_depends-v5.9.patch > new file mode 100644 > index 0000000000..0970dd30aa > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0006-fix-removal-of-smp_-read_barrier_depends-v5.9.patch > @@ -0,0 +1,391 @@ > +From 57ccbfa6a8a79c7b84394c2097efaf7935607aa5 Mon Sep 17 00:00:00 2001 > +From: Michael Jeanson <[email protected]> > +Date: Tue, 25 Aug 2020 10:56:29 -0400 > +Subject: [PATCH 06/10] fix: removal of [smp_]read_barrier_depends (v5.9) > + > +See upstream commits: > + > + commit 76ebbe78f7390aee075a7f3768af197ded1bdfbb > + Author: Will Deacon <[email protected]> > + Date: Tue Oct 24 11:22:47 2017 +0100 > + > + locking/barriers: Add implicit smp_read_barrier_depends() to READ_ONCE() > + > + In preparation for the removal of lockless_dereference(), which is the > + same as READ_ONCE() on all architectures other than Alpha, add an > + implicit smp_read_barrier_depends() to READ_ONCE() so that it can be > + used to head dependency chains on all architectures. > + > + commit 76ebbe78f7390aee075a7f3768af197ded1bdfbb > + Author: Will Deacon <[email protected]> > + Date: Tue Oct 24 11:22:47 2017 +0100 > + > + locking/barriers: Add implicit smp_read_barrier_depends() to READ_ONCE() > + > + In preparation for the removal of lockless_dereference(), which is the > + same as READ_ONCE() on all architectures other than Alpha, add an > + implicit smp_read_barrier_depends() to READ_ONCE() so that it can be > + used to head dependency chains on all architectures. > + > +Upstream-Status: Backport > + > +Change-Id: Ife8880bd9378dca2972da8838f40fc35ccdfaaac > +Signed-off-by: Michael Jeanson <[email protected]> > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +--- > + instrumentation/events/lttng-module/i2c.h | 4 ++-- > + lib/ringbuffer/backend.h | 2 +- > + lib/ringbuffer/backend_internal.h | 2 +- > + lib/ringbuffer/frontend.h | 4 ++-- > + lib/ringbuffer/ring_buffer_frontend.c | 4 ++-- > + lib/ringbuffer/ring_buffer_iterator.c | 2 +- > + lttng-events.c | 8 ++++---- > + probes/lttng-kprobes.c | 6 +++--- > + probes/lttng-kretprobes.c | 6 +++--- > + probes/lttng-tracepoint-event-impl.h | 12 ++++++------ > + probes/lttng-uprobes.c | 6 +++--- > + wrapper/compiler.h | 18 ++++++++++++++++++ > + wrapper/trace-clock.h | 15 +++++---------- > + 13 files changed, 51 insertions(+), 38 deletions(-) > + > +diff --git a/instrumentation/events/lttng-module/i2c.h > b/instrumentation/events/lttng-module/i2c.h > +index dcbabf6..131d134 100644 > +--- a/instrumentation/events/lttng-module/i2c.h > ++++ b/instrumentation/events/lttng-module/i2c.h > +@@ -23,7 +23,7 @@ LTTNG_TRACEPOINT_EVENT_CODE(i2c_write, > + > + TP_code_pre( > + tp_locvar->extract_sensitive_payload = > +- READ_ONCE(extract_sensitive_payload); > ++ LTTNG_READ_ONCE(extract_sensitive_payload); > + ), > + > + TP_FIELDS( > +@@ -78,7 +78,7 @@ LTTNG_TRACEPOINT_EVENT_CODE(i2c_reply, > + > + TP_code_pre( > + tp_locvar->extract_sensitive_payload = > +- READ_ONCE(extract_sensitive_payload); > ++ LTTNG_READ_ONCE(extract_sensitive_payload); > + ), > + > + TP_FIELDS( > +diff --git a/lib/ringbuffer/backend.h b/lib/ringbuffer/backend.h > +index da937f2..43e1d47 100644 > +--- a/lib/ringbuffer/backend.h > ++++ b/lib/ringbuffer/backend.h > +@@ -156,7 +156,7 @@ size_t lib_ring_buffer_do_strcpy(const struct > lib_ring_buffer_config *config, > + * Only read source character once, in case it is > + * modified concurrently. > + */ > +- c = READ_ONCE(src[count]); > ++ c = LTTNG_READ_ONCE(src[count]); > + if (!c) > + break; > + lib_ring_buffer_do_copy(config, &dest[count], &c, 1); > +diff --git a/lib/ringbuffer/backend_internal.h > b/lib/ringbuffer/backend_internal.h > +index 2d6a345..1226fd8 100644 > +--- a/lib/ringbuffer/backend_internal.h > ++++ b/lib/ringbuffer/backend_internal.h > +@@ -367,7 +367,7 @@ void lib_ring_buffer_clear_noref(const struct > lib_ring_buffer_config *config, > + * Performing a volatile access to read the sb_pages, because we want > to > + * read a coherent version of the pointer and the associated noref > flag. > + */ > +- id = READ_ONCE(bufb->buf_wsb[idx].id); > ++ id = LTTNG_READ_ONCE(bufb->buf_wsb[idx].id); > + for (;;) { > + /* This check is called on the fast path for each record. */ > + if (likely(!subbuffer_id_is_noref(config, id))) { > +diff --git a/lib/ringbuffer/frontend.h b/lib/ringbuffer/frontend.h > +index 6f516d9..41382fe 100644 > +--- a/lib/ringbuffer/frontend.h > ++++ b/lib/ringbuffer/frontend.h > +@@ -79,7 +79,7 @@ void *channel_destroy(struct channel *chan); > + #define for_each_channel_cpu(cpu, chan) > \ > + for ((cpu) = -1; \ > + ({ (cpu) = cpumask_next(cpu, (chan)->backend.cpumask); \ > +- smp_read_barrier_depends(); (cpu) < nr_cpu_ids; });) > ++ smp_rmb(); (cpu) < nr_cpu_ids; });) > + > + extern struct lib_ring_buffer *channel_get_ring_buffer( > + const struct lib_ring_buffer_config *config, > +@@ -155,7 +155,7 @@ static inline > + int lib_ring_buffer_is_finalized(const struct lib_ring_buffer_config > *config, > + struct lib_ring_buffer *buf) > + { > +- int finalized = READ_ONCE(buf->finalized); > ++ int finalized = LTTNG_READ_ONCE(buf->finalized); > + /* > + * Read finalized before counters. > + */ > +diff --git a/lib/ringbuffer/ring_buffer_frontend.c > b/lib/ringbuffer/ring_buffer_frontend.c > +index 3cab365..4980d20 100644 > +--- a/lib/ringbuffer/ring_buffer_frontend.c > ++++ b/lib/ringbuffer/ring_buffer_frontend.c > +@@ -1074,7 +1074,7 @@ int lib_ring_buffer_snapshot(struct lib_ring_buffer > *buf, > + int finalized; > + > + retry: > +- finalized = READ_ONCE(buf->finalized); > ++ finalized = LTTNG_READ_ONCE(buf->finalized); > + /* > + * Read finalized before counters. > + */ > +@@ -1245,7 +1245,7 @@ int lib_ring_buffer_get_subbuf(struct lib_ring_buffer > *buf, > + return -EBUSY; > + } > + retry: > +- finalized = READ_ONCE(buf->finalized); > ++ finalized = LTTNG_READ_ONCE(buf->finalized); > + /* > + * Read finalized before counters. > + */ > +diff --git a/lib/ringbuffer/ring_buffer_iterator.c > b/lib/ringbuffer/ring_buffer_iterator.c > +index d25db72..7b4f20a 100644 > +--- a/lib/ringbuffer/ring_buffer_iterator.c > ++++ b/lib/ringbuffer/ring_buffer_iterator.c > +@@ -46,7 +46,7 @@ restart: > + switch (iter->state) { > + case ITER_GET_SUBBUF: > + ret = lib_ring_buffer_get_next_subbuf(buf); > +- if (ret && !READ_ONCE(buf->finalized) > ++ if (ret && !LTTNG_READ_ONCE(buf->finalized) > + && config->alloc == RING_BUFFER_ALLOC_GLOBAL) { > + /* > + * Use "pull" scheme for global buffers. The reader > +diff --git a/lttng-events.c b/lttng-events.c > +index be7e389..d719294 100644 > +--- a/lttng-events.c > ++++ b/lttng-events.c > +@@ -1719,7 +1719,7 @@ int lttng_metadata_printf(struct lttng_session > *session, > + size_t len; > + va_list ap; > + > +- WARN_ON_ONCE(!READ_ONCE(session->active)); > ++ WARN_ON_ONCE(!LTTNG_READ_ONCE(session->active)); > + > + va_start(ap, fmt); > + str = kvasprintf(GFP_KERNEL, fmt, ap); > +@@ -2305,7 +2305,7 @@ int _lttng_event_metadata_statedump(struct > lttng_session *session, > + { > + int ret = 0; > + > +- if (event->metadata_dumped || !READ_ONCE(session->active)) > ++ if (event->metadata_dumped || !LTTNG_READ_ONCE(session->active)) > + return 0; > + if (chan->channel_type == METADATA_CHANNEL) > + return 0; > +@@ -2377,7 +2377,7 @@ int _lttng_channel_metadata_statedump(struct > lttng_session *session, > + { > + int ret = 0; > + > +- if (chan->metadata_dumped || !READ_ONCE(session->active)) > ++ if (chan->metadata_dumped || !LTTNG_READ_ONCE(session->active)) > + return 0; > + > + if (chan->channel_type == METADATA_CHANNEL) > +@@ -2604,7 +2604,7 @@ int _lttng_session_metadata_statedump(struct > lttng_session *session) > + struct lttng_event *event; > + int ret = 0; > + > +- if (!READ_ONCE(session->active)) > ++ if (!LTTNG_READ_ONCE(session->active)) > + return 0; > + > + lttng_metadata_begin(session); > +diff --git a/probes/lttng-kprobes.c b/probes/lttng-kprobes.c > +index a44eaa1..38fb72e 100644 > +--- a/probes/lttng-kprobes.c > ++++ b/probes/lttng-kprobes.c > +@@ -31,11 +31,11 @@ int lttng_kprobes_handler_pre(struct kprobe *p, struct > pt_regs *regs) > + int ret; > + unsigned long data = (unsigned long) p->addr; > + > +- if (unlikely(!READ_ONCE(chan->session->active))) > ++ if (unlikely(!LTTNG_READ_ONCE(chan->session->active))) > + return 0; > +- if (unlikely(!READ_ONCE(chan->enabled))) > ++ if (unlikely(!LTTNG_READ_ONCE(chan->enabled))) > + return 0; > +- if (unlikely(!READ_ONCE(event->enabled))) > ++ if (unlikely(!LTTNG_READ_ONCE(event->enabled))) > + return 0; > + > + lib_ring_buffer_ctx_init(&ctx, chan->chan, <tng_probe_ctx, > sizeof(data), > +diff --git a/probes/lttng-kretprobes.c b/probes/lttng-kretprobes.c > +index ab98ff2..a6bcd21 100644 > +--- a/probes/lttng-kretprobes.c > ++++ b/probes/lttng-kretprobes.c > +@@ -51,11 +51,11 @@ int _lttng_kretprobes_handler(struct kretprobe_instance > *krpi, > + unsigned long parent_ip; > + } payload; > + > +- if (unlikely(!READ_ONCE(chan->session->active))) > ++ if (unlikely(!LTTNG_READ_ONCE(chan->session->active))) > + return 0; > +- if (unlikely(!READ_ONCE(chan->enabled))) > ++ if (unlikely(!LTTNG_READ_ONCE(chan->enabled))) > + return 0; > +- if (unlikely(!READ_ONCE(event->enabled))) > ++ if (unlikely(!LTTNG_READ_ONCE(event->enabled))) > + return 0; > + > + payload.ip = (unsigned long) krpi->rp->kp.addr; > +diff --git a/probes/lttng-tracepoint-event-impl.h > b/probes/lttng-tracepoint-event-impl.h > +index 77b8638..72a669e 100644 > +--- a/probes/lttng-tracepoint-event-impl.h > ++++ b/probes/lttng-tracepoint-event-impl.h > +@@ -1132,11 +1132,11 @@ static void __event_probe__##_name(void *__data, > _proto) \ > + > \ > + if (!_TP_SESSION_CHECK(session, __session)) > \ > + return; > \ > +- if (unlikely(!READ_ONCE(__session->active))) > \ > ++ if (unlikely(!LTTNG_READ_ONCE(__session->active))) > \ > + return; > \ > +- if (unlikely(!READ_ONCE(__chan->enabled))) > \ > ++ if (unlikely(!LTTNG_READ_ONCE(__chan->enabled))) > \ > + return; > \ > +- if (unlikely(!READ_ONCE(__event->enabled))) > \ > ++ if (unlikely(!LTTNG_READ_ONCE(__event->enabled))) > \ > + return; > \ > + __lf = lttng_rcu_dereference(__session->pid_tracker.p); > \ > + if (__lf && likely(!lttng_id_tracker_lookup(__lf, current->tgid))) > \ > +@@ -1225,11 +1225,11 @@ static void __event_probe__##_name(void *__data) > \ > + > \ > + if (!_TP_SESSION_CHECK(session, __session)) > \ > + return; > \ > +- if (unlikely(!READ_ONCE(__session->active))) > \ > ++ if (unlikely(!LTTNG_READ_ONCE(__session->active))) > \ > + return; > \ > +- if (unlikely(!READ_ONCE(__chan->enabled))) > \ > ++ if (unlikely(!LTTNG_READ_ONCE(__chan->enabled))) > \ > + return; > \ > +- if (unlikely(!READ_ONCE(__event->enabled))) > \ > ++ if (unlikely(!LTTNG_READ_ONCE(__event->enabled))) > \ > + return; > \ > + __lf = lttng_rcu_dereference(__session->pid_tracker.p); > \ > + if (__lf && likely(!lttng_id_tracker_lookup(__lf, current->tgid))) > \ > +diff --git a/probes/lttng-uprobes.c b/probes/lttng-uprobes.c > +index bc10128..bda1d9b 100644 > +--- a/probes/lttng-uprobes.c > ++++ b/probes/lttng-uprobes.c > +@@ -40,11 +40,11 @@ int lttng_uprobes_handler_pre(struct uprobe_consumer > *uc, struct pt_regs *regs) > + unsigned long ip; > + } payload; > + > +- if (unlikely(!READ_ONCE(chan->session->active))) > ++ if (unlikely(!LTTNG_READ_ONCE(chan->session->active))) > + return 0; > +- if (unlikely(!READ_ONCE(chan->enabled))) > ++ if (unlikely(!LTTNG_READ_ONCE(chan->enabled))) > + return 0; > +- if (unlikely(!READ_ONCE(event->enabled))) > ++ if (unlikely(!LTTNG_READ_ONCE(event->enabled))) > + return 0; > + > + lib_ring_buffer_ctx_init(&ctx, chan->chan, <tng_probe_ctx, > +diff --git a/wrapper/compiler.h b/wrapper/compiler.h > +index 1496f33..b9f8c51 100644 > +--- a/wrapper/compiler.h > ++++ b/wrapper/compiler.h > +@@ -9,6 +9,7 @@ > + #define _LTTNG_WRAPPER_COMPILER_H > + > + #include <linux/compiler.h> > ++#include <linux/version.h> > + > + /* > + * Don't allow compiling with buggy compiler. > +@@ -39,4 +40,21 @@ > + # define WRITE_ONCE(x, val) ({ ACCESS_ONCE(x) = val; }) > + #endif > + > ++/* > ++ * In v4.15 a smp read barrier was added to READ_ONCE to replace > ++ * lockless_dereference(), replicate this behavior on prior kernels > ++ * and remove calls to smp_read_barrier_depends which was dropped > ++ * in v5.9. > ++ */ > ++#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,15,0)) > ++#define LTTNG_READ_ONCE(x) READ_ONCE(x) > ++#else > ++#define LTTNG_READ_ONCE(x) \ > ++({ \ > ++ typeof(x) __val = READ_ONCE(x); \ > ++ smp_read_barrier_depends(); \ > ++ __val; \ > ++}) > ++#endif > ++ > + #endif /* _LTTNG_WRAPPER_COMPILER_H */ > +diff --git a/wrapper/trace-clock.h b/wrapper/trace-clock.h > +index 9f4e366..187fc82 100644 > +--- a/wrapper/trace-clock.h > ++++ b/wrapper/trace-clock.h > +@@ -160,33 +160,30 @@ static inline void put_trace_clock(void) > + > + static inline u64 trace_clock_read64(void) > + { > +- struct lttng_trace_clock *ltc = READ_ONCE(lttng_trace_clock); > ++ struct lttng_trace_clock *ltc = LTTNG_READ_ONCE(lttng_trace_clock); > + > + if (likely(!ltc)) { > + return trace_clock_read64_monotonic(); > + } else { > +- read_barrier_depends(); /* load ltc before content */ > + return ltc->read64(); > + } > + } > + > + static inline u64 trace_clock_freq(void) > + { > +- struct lttng_trace_clock *ltc = READ_ONCE(lttng_trace_clock); > ++ struct lttng_trace_clock *ltc = LTTNG_READ_ONCE(lttng_trace_clock); > + > + if (!ltc) { > + return trace_clock_freq_monotonic(); > + } else { > +- read_barrier_depends(); /* load ltc before content */ > + return ltc->freq(); > + } > + } > + > + static inline int trace_clock_uuid(char *uuid) > + { > +- struct lttng_trace_clock *ltc = READ_ONCE(lttng_trace_clock); > ++ struct lttng_trace_clock *ltc = LTTNG_READ_ONCE(lttng_trace_clock); > + > +- read_barrier_depends(); /* load ltc before content */ > + /* Use default UUID cb when NULL */ > + if (!ltc || !ltc->uuid) { > + return trace_clock_uuid_monotonic(uuid); > +@@ -197,24 +194,22 @@ static inline int trace_clock_uuid(char *uuid) > + > + static inline const char *trace_clock_name(void) > + { > +- struct lttng_trace_clock *ltc = READ_ONCE(lttng_trace_clock); > ++ struct lttng_trace_clock *ltc = LTTNG_READ_ONCE(lttng_trace_clock); > + > + if (!ltc) { > + return trace_clock_name_monotonic(); > + } else { > +- read_barrier_depends(); /* load ltc before content */ > + return ltc->name(); > + } > + } > + > + static inline const char *trace_clock_description(void) > + { > +- struct lttng_trace_clock *ltc = READ_ONCE(lttng_trace_clock); > ++ struct lttng_trace_clock *ltc = LTTNG_READ_ONCE(lttng_trace_clock); > + > + if (!ltc) { > + return trace_clock_description_monotonic(); > + } else { > +- read_barrier_depends(); /* load ltc before content */ > + return ltc->description(); > + } > + } > +-- > +2.19.1 > + > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0007-fix-writeback-Drop-I_DIRTY_TIME_EXPIRE-v5.9.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0007-fix-writeback-Drop-I_DIRTY_TIME_EXPIRE-v5.9.patch > new file mode 100644 > index 0000000000..2843c9cb62 > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0007-fix-writeback-Drop-I_DIRTY_TIME_EXPIRE-v5.9.patch > @@ -0,0 +1,59 @@ > +From eae02feb58064eee5ce15a9f6bdffd107c47da05 Mon Sep 17 00:00:00 2001 > +From: Michael Jeanson <[email protected]> > +Date: Mon, 31 Aug 2020 11:41:38 -0400 > +Subject: [PATCH 07/10] fix: writeback: Drop I_DIRTY_TIME_EXPIRE (v5.9) > + > +See upstream commit: > + > + commit 5fcd57505c002efc5823a7355e21f48dd02d5a51 > + Author: Jan Kara <[email protected]> > + Date: Fri May 29 16:24:43 2020 +0200 > + > + writeback: Drop I_DIRTY_TIME_EXPIRE > + > + The only use of I_DIRTY_TIME_EXPIRE is to detect in > + __writeback_single_inode() that inode got there because flush worker > + decided it's time to writeback the dirty inode time stamps (either > + because we are syncing or because of age). However we can detect this > + directly in __writeback_single_inode() and there's no need for the > + strange propagation with I_DIRTY_TIME_EXPIRE flag. > + > +Upstream-Status: Backport > + > +Signed-off-by: Michael Jeanson <[email protected]> > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +Change-Id: I92e37c2ff3ec36d431e8f9de5c8e37c5a2da55ea > +--- > + instrumentation/events/lttng-module/writeback.h | 16 +++++++++++++++- > + 1 file changed, 15 insertions(+), 1 deletion(-) > + > +diff --git a/instrumentation/events/lttng-module/writeback.h > b/instrumentation/events/lttng-module/writeback.h > +index affb4eb..ece67ad 100644 > +--- a/instrumentation/events/lttng-module/writeback.h > ++++ b/instrumentation/events/lttng-module/writeback.h > +@@ -46,7 +46,21 @@ static inline struct backing_dev_info > *lttng_inode_to_bdi(struct inode *inode) > + > + #endif > + > +-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0)) > ++#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0)) > ++#define show_inode_state(state) \ > ++ __print_flags(state, "|", \ > ++ {I_DIRTY_SYNC, "I_DIRTY_SYNC"}, \ > ++ {I_DIRTY_DATASYNC, "I_DIRTY_DATASYNC"}, \ > ++ {I_DIRTY_PAGES, "I_DIRTY_PAGES"}, \ > ++ {I_NEW, "I_NEW"}, \ > ++ {I_WILL_FREE, "I_WILL_FREE"}, \ > ++ {I_FREEING, "I_FREEING"}, \ > ++ {I_CLEAR, "I_CLEAR"}, \ > ++ {I_SYNC, "I_SYNC"}, \ > ++ {I_DIRTY_TIME, "I_DIRTY_TIME"}, \ > ++ {I_REFERENCED, "I_REFERENCED"} \ > ++ ) > ++#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(4,0,0)) > + #define show_inode_state(state) \ > + __print_flags(state, "|", \ > + {I_DIRTY_SYNC, "I_DIRTY_SYNC"}, \ > +-- > +2.19.1 > + > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0008-fix-writeback-Fix-sync-livelock-due-to-b_dirty_time-.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0008-fix-writeback-Fix-sync-livelock-due-to-b_dirty_time-.patch > new file mode 100644 > index 0000000000..7a0d9a38b8 > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0008-fix-writeback-Fix-sync-livelock-due-to-b_dirty_time-.patch > @@ -0,0 +1,117 @@ > +From 87b2affc3eb06f3fb2d0923f18af37713eb6814b Mon Sep 17 00:00:00 2001 > +From: Michael Jeanson <[email protected]> > +Date: Mon, 31 Aug 2020 14:16:01 -0400 > +Subject: [PATCH 08/10] fix: writeback: Fix sync livelock due to b_dirty_time > + processing (v5.9) > + > +See upstream commit: > + > + commit f9cae926f35e8230330f28c7b743ad088611a8de > + Author: Jan Kara <[email protected]> > + Date: Fri May 29 16:08:58 2020 +0200 > + > + writeback: Fix sync livelock due to b_dirty_time processing > + > + When we are processing writeback for sync(2), move_expired_inodes() > + didn't set any inode expiry value (older_than_this). This can result in > + writeback never completing if there's steady stream of inodes added to > + b_dirty_time list as writeback rechecks dirty lists after each writeback > + round whether there's more work to be done. Fix the problem by using > + sync(2) start time is inode expiry value when processing b_dirty_time > + list similarly as for ordinarily dirtied inodes. This requires some > + refactoring of older_than_this handling which simplifies the code > + noticeably as a bonus. > + > +Upstream-Status: Backport > + > +Change-Id: I8b894b13ccc14d9b8983ee4c2810a927c319560b > +Signed-off-by: Michael Jeanson <[email protected]> > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +--- > + .../events/lttng-module/writeback.h | 39 ++++++++++++------- > + 1 file changed, 26 insertions(+), 13 deletions(-) > + > +diff --git a/instrumentation/events/lttng-module/writeback.h > b/instrumentation/events/lttng-module/writeback.h > +index ece67ad..e9018dd 100644 > +--- a/instrumentation/events/lttng-module/writeback.h > ++++ b/instrumentation/events/lttng-module/writeback.h > +@@ -384,34 +384,48 @@ > LTTNG_TRACEPOINT_EVENT_WBC_INSTANCE(wbc_balance_dirty_wait, > writeback_wbc_balanc > + #endif > + LTTNG_TRACEPOINT_EVENT_WBC_INSTANCE(wbc_writepage, writeback_wbc_writepage) > + > +-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,1,0)) > ++#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0)) > ++LTTNG_TRACEPOINT_EVENT(writeback_queue_io, > ++ TP_PROTO(struct bdi_writeback *wb, > ++ struct wb_writeback_work *work, > ++ unsigned long dirtied_before, > ++ int moved), > ++ TP_ARGS(wb, work, dirtied_before, moved), > ++ TP_FIELDS( > ++ ctf_array_text(char, name, dev_name(wb->bdi->dev), 32) > ++ ctf_integer(unsigned long, older, dirtied_before) > ++ ctf_integer(int, moved, moved) > ++ ) > ++) > ++#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0)) > + LTTNG_TRACEPOINT_EVENT(writeback_queue_io, > + TP_PROTO(struct bdi_writeback *wb, > +-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0)) > + struct wb_writeback_work *work, > +-#else > +- unsigned long *older_than_this, > +-#endif > + int moved), > +-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0)) > + TP_ARGS(wb, work, moved), > +-#else > ++ TP_FIELDS( > ++ ctf_array_text(char, name, dev_name(wb->bdi->dev), 32) > ++ ctf_integer(int, moved, moved) > ++ ) > ++) > ++#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,1,0)) > ++LTTNG_TRACEPOINT_EVENT(writeback_queue_io, > ++ TP_PROTO(struct bdi_writeback *wb, > ++ unsigned long *older_than_this, > ++ int moved), > + TP_ARGS(wb, older_than_this, moved), > +-#endif > + TP_FIELDS( > + ctf_array_text(char, name, dev_name(wb->bdi->dev), 32) > +-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0)) > +-#else > + ctf_integer(unsigned long, older, > + older_than_this ? *older_than_this : 0) > + ctf_integer(long, age, > + older_than_this ? > + (jiffies - *older_than_this) * 1000 / HZ > + : -1) > +-#endif > + ctf_integer(int, moved, moved) > + ) > + ) > ++#endif > + > + #if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,8,0)) > + LTTNG_TRACEPOINT_EVENT_MAP(global_dirty_state, > +@@ -460,7 +474,7 @@ LTTNG_TRACEPOINT_EVENT_MAP(global_dirty_state, > + ctf_integer(unsigned long, dirty_limit, global_dirty_limit) > + ) > + ) > +-#else > ++#elif (LINUX_VERSION_CODE >= KERNEL_VERSION(3,1,0)) > + LTTNG_TRACEPOINT_EVENT_MAP(global_dirty_state, > + > + writeback_global_dirty_state, > +@@ -485,7 +499,6 @@ LTTNG_TRACEPOINT_EVENT_MAP(global_dirty_state, > + ) > + ) > + #endif > +-#endif > + > + #if (LINUX_VERSION_CODE >= KERNEL_VERSION(3,2,0)) > + > +-- > +2.19.1 > + > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0009-fix-version-ranges-for-ext4_discard_preallocations-a.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0009-fix-version-ranges-for-ext4_discard_preallocations-a.patch > new file mode 100644 > index 0000000000..346e1d63ad > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0009-fix-version-ranges-for-ext4_discard_preallocations-a.patch > @@ -0,0 +1,52 @@ > +From b74b25f349e92d7b5bdc8684e406d6a889f13773 Mon Sep 17 00:00:00 2001 > +From: Michael Jeanson <[email protected]> > +Date: Fri, 4 Sep 2020 11:52:51 -0400 > +Subject: [PATCH 09/10] fix: version ranges for ext4_discard_preallocations > and > + writeback_queue_io > + > +Upstream-Status: Backport > + > +Signed-off-by: Michael Jeanson <[email protected]> > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +Change-Id: Id4fa53cb2e713cbda651e1a75deed91013115592 > +--- > + instrumentation/events/lttng-module/ext4.h | 3 ++- > + instrumentation/events/lttng-module/writeback.h | 8 +++++++- > + 2 files changed, 9 insertions(+), 2 deletions(-) > + > +diff --git a/instrumentation/events/lttng-module/ext4.h > b/instrumentation/events/lttng-module/ext4.h > +index 4476abb..b172c8d 100644 > +--- a/instrumentation/events/lttng-module/ext4.h > ++++ b/instrumentation/events/lttng-module/ext4.h > +@@ -460,7 +460,8 @@ LTTNG_TRACEPOINT_EVENT(ext4_mb_release_group_pa, > + ) > + #endif > + > +-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0)) > ++#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0) || \ > ++ LTTNG_KERNEL_RANGE(5,8,6, 5,9,0)) > + LTTNG_TRACEPOINT_EVENT(ext4_discard_preallocations, > + TP_PROTO(struct inode *inode, unsigned int len, unsigned int needed), > + > +diff --git a/instrumentation/events/lttng-module/writeback.h > b/instrumentation/events/lttng-module/writeback.h > +index e9018dd..09637d7 100644 > +--- a/instrumentation/events/lttng-module/writeback.h > ++++ b/instrumentation/events/lttng-module/writeback.h > +@@ -384,7 +384,13 @@ > LTTNG_TRACEPOINT_EVENT_WBC_INSTANCE(wbc_balance_dirty_wait, > writeback_wbc_balanc > + #endif > + LTTNG_TRACEPOINT_EVENT_WBC_INSTANCE(wbc_writepage, writeback_wbc_writepage) > + > +-#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0)) > ++#if (LINUX_VERSION_CODE >= KERNEL_VERSION(5,9,0) || \ > ++ LTTNG_KERNEL_RANGE(5,8,6, 5,9,0) || \ > ++ LTTNG_KERNEL_RANGE(5,4,62, 5,5,0) || \ > ++ LTTNG_KERNEL_RANGE(4,19,143, 4,20,0) || \ > ++ LTTNG_KERNEL_RANGE(4,14,196, 4,15,0) || \ > ++ LTTNG_KERNEL_RANGE(4,9,235, 4,10,0) || \ > ++ LTTNG_KERNEL_RANGE(4,4,235, 4,5,0)) > + LTTNG_TRACEPOINT_EVENT(writeback_queue_io, > + TP_PROTO(struct bdi_writeback *wb, > + struct wb_writeback_work *work, > +-- > +2.19.1 > + > diff --git > a/meta/recipes-kernel/lttng/lttng-modules/0010-Fix-system-call-filter-table.patch > > b/meta/recipes-kernel/lttng/lttng-modules/0010-Fix-system-call-filter-table.patch > new file mode 100644 > index 0000000000..a16750ddb3 > --- /dev/null > +++ > b/meta/recipes-kernel/lttng/lttng-modules/0010-Fix-system-call-filter-table.patch > @@ -0,0 +1,918 @@ > +From ad594e3a953db1b0c3c059fde45b5a5494f6be78 Mon Sep 17 00:00:00 2001 > +From: Mathieu Desnoyers <[email protected]> > +Date: Tue, 28 Jan 2020 16:02:44 -0500 > +Subject: [PATCH 10/10] Fix: system call filter table > + > +The system call filter table has effectively been unused for a long > +time due to system call name prefix mismatch. This means the overhead of > +selective system call tracing was larger than it should have been because > +the event payload preparation would be done for all system calls as soon > +as a single system call is traced. > + > +However, fixing this underlying issue unearths several issues that crept > +unnoticed when the "enabler" concept was introduced (after the original > +implementation of the system call filter table). > + > +Here is a list of the issues which are resolved here: > + > +- Split lttng_syscalls_unregister into an unregister and destroy > + function, thus awaiting for a grace period (and therefore quiescence > + of the users) after unregistering the system call tracepoints before > + freeing the system call filter data structures. This effectively fixes > + a use-after-free. > + > +- The state for enabling "all" system calls vs enabling specific system > + calls (and sequences of enable-disable) was incorrect with respect to > + the "enablers" semantic. This is solved by always tracking the > + bitmap of enabled system calls, and keeping this bitmap even when > + enabling all system calls. The sc_filter is now always allocated > + before system call tracing is registered to tracepoints, which means > + it does not need to be RCU dereferenced anymore. > + > +Padding fields in the ABI are reserved to select whether to: > + > +- Trace either native or compat system call (or both, which is the > + behavior currently implemented), > +- Trace either system call entry or exit (or both, which is the > + behavior currently implemented), > +- Select the system call to trace by name (behavior currently > + implemented) or by system call number, > + > +Upstream-Status: Backport > + > +Signed-off-by: Mathieu Desnoyers <[email protected]> > +--- > + lttng-abi.c | 43 ++++++ > + lttng-abi.h | 26 ++++ > + lttng-events.c | 112 +++++++++++++-- > + lttng-events.h | 31 ++++- > + lttng-syscalls.c | 348 +++++++++++++++++++++++++---------------------- > + 5 files changed, 380 insertions(+), 180 deletions(-) > + > +diff --git a/lttng-abi.c b/lttng-abi.c > +index 64ea99d..b33879d 100644 > +--- a/lttng-abi.c > ++++ b/lttng-abi.c > +@@ -1264,6 +1264,46 @@ nomem: > + return ret; > + } > + > ++static > ++int lttng_abi_validate_event_param(struct lttng_kernel_event *event_param) > ++{ > ++ /* Limit ABI to implemented features. */ > ++ switch (event_param->instrumentation) { > ++ case LTTNG_KERNEL_SYSCALL: > ++ switch (event_param->u.syscall.entryexit) { > ++ case LTTNG_KERNEL_SYSCALL_ENTRYEXIT: > ++ break; > ++ default: > ++ return -EINVAL; > ++ } > ++ switch (event_param->u.syscall.abi) { > ++ case LTTNG_KERNEL_SYSCALL_ABI_ALL: > ++ break; > ++ default: > ++ return -EINVAL; > ++ } > ++ switch (event_param->u.syscall.match) { > ++ case LTTNG_SYSCALL_MATCH_NAME: > ++ break; > ++ default: > ++ return -EINVAL; > ++ } > ++ break; > ++ > ++ case LTTNG_KERNEL_TRACEPOINT: /* Fallthrough */ > ++ case LTTNG_KERNEL_KPROBE: /* Fallthrough */ > ++ case LTTNG_KERNEL_KRETPROBE: /* Fallthrough */ > ++ case LTTNG_KERNEL_NOOP: /* Fallthrough */ > ++ case LTTNG_KERNEL_UPROBE: > ++ break; > ++ > ++ case LTTNG_KERNEL_FUNCTION: /* Fallthrough */ > ++ default: > ++ return -EINVAL; > ++ } > ++ return 0; > ++} > ++ > + static > + int lttng_abi_create_event(struct file *channel_file, > + struct lttng_kernel_event *event_param) > +@@ -1305,6 +1345,9 @@ int lttng_abi_create_event(struct file *channel_file, > + ret = -EOVERFLOW; > + goto refcount_error; > + } > ++ ret = lttng_abi_validate_event_param(event_param); > ++ if (ret) > ++ goto event_error; > + if (event_param->instrumentation == LTTNG_KERNEL_TRACEPOINT > + || event_param->instrumentation == > LTTNG_KERNEL_SYSCALL) { > + struct lttng_enabler *enabler; > +diff --git a/lttng-abi.h b/lttng-abi.h > +index 1d356ab..51d60e5 100644 > +--- a/lttng-abi.h > ++++ b/lttng-abi.h > +@@ -90,6 +90,31 @@ struct lttng_kernel_event_callsite { > + } u; > + } __attribute__((packed)); > + > ++enum lttng_kernel_syscall_entryexit { > ++ LTTNG_KERNEL_SYSCALL_ENTRYEXIT = 0, > ++ LTTNG_KERNEL_SYSCALL_ENTRY = 1, /* Not implemented. */ > ++ LTTNG_KERNEL_SYSCALL_EXIT = 2, /* Not implemented. */ > ++}; > ++ > ++enum lttng_kernel_syscall_abi { > ++ LTTNG_KERNEL_SYSCALL_ABI_ALL = 0, > ++ LTTNG_KERNEL_SYSCALL_ABI_NATIVE = 1, /* Not implemented. */ > ++ LTTNG_KERNEL_SYSCALL_ABI_COMPAT = 2, /* Not implemented. */ > ++}; > ++ > ++enum lttng_kernel_syscall_match { > ++ LTTNG_SYSCALL_MATCH_NAME = 0, > ++ LTTNG_SYSCALL_MATCH_NR = 1, /* Not implemented. */ > ++}; > ++ > ++struct lttng_kernel_syscall { > ++ uint8_t entryexit; /* enum lttng_kernel_syscall_entryexit */ > ++ uint8_t abi; /* enum lttng_kernel_syscall_abi */ > ++ uint8_t match; /* enum lttng_kernel_syscall_match */ > ++ uint8_t padding; > ++ uint32_t nr; /* For LTTNG_SYSCALL_MATCH_NR */ > ++} __attribute__((packed)); > ++ > + /* > + * For syscall tracing, name = "*" means "enable all". > + */ > +@@ -106,6 +131,7 @@ struct lttng_kernel_event { > + struct lttng_kernel_kprobe kprobe; > + struct lttng_kernel_function_tracer ftrace; > + struct lttng_kernel_uprobe uprobe; > ++ struct lttng_kernel_syscall syscall; > + char padding[LTTNG_KERNEL_EVENT_PADDING2]; > + } u; > + } __attribute__((packed)); > +diff --git a/lttng-events.c b/lttng-events.c > +index d719294..4c0b04a 100644 > +--- a/lttng-events.c > ++++ b/lttng-events.c > +@@ -201,6 +201,10 @@ void lttng_session_destroy(struct lttng_session > *session) > + WARN_ON(ret); > + } > + synchronize_trace(); /* Wait for in-flight events to complete */ > ++ list_for_each_entry(chan, &session->chan, list) { > ++ ret = lttng_syscalls_destroy(chan); > ++ WARN_ON(ret); > ++ } > + list_for_each_entry_safe(enabler, tmpenabler, > + &session->enablers_head, node) > + lttng_enabler_destroy(enabler); > +@@ -740,6 +744,28 @@ struct lttng_event *_lttng_event_create(struct > lttng_channel *chan, > + event->enabled = 0; > + event->registered = 0; > + event->desc = event_desc; > ++ switch (event_param->u.syscall.entryexit) { > ++ case LTTNG_KERNEL_SYSCALL_ENTRYEXIT: > ++ ret = -EINVAL; > ++ goto register_error; > ++ case LTTNG_KERNEL_SYSCALL_ENTRY: > ++ event->u.syscall.entryexit = LTTNG_SYSCALL_ENTRY; > ++ break; > ++ case LTTNG_KERNEL_SYSCALL_EXIT: > ++ event->u.syscall.entryexit = LTTNG_SYSCALL_EXIT; > ++ break; > ++ } > ++ switch (event_param->u.syscall.abi) { > ++ case LTTNG_KERNEL_SYSCALL_ABI_ALL: > ++ ret = -EINVAL; > ++ goto register_error; > ++ case LTTNG_KERNEL_SYSCALL_ABI_NATIVE: > ++ event->u.syscall.abi = LTTNG_SYSCALL_ABI_NATIVE; > ++ break; > ++ case LTTNG_KERNEL_SYSCALL_ABI_COMPAT: > ++ event->u.syscall.abi = LTTNG_SYSCALL_ABI_COMPAT; > ++ break; > ++ } > + if (!event->desc) { > + ret = -EINVAL; > + goto register_error; > +@@ -826,8 +852,7 @@ void register_event(struct lttng_event *event) > + event); > + break; > + case LTTNG_KERNEL_SYSCALL: > +- ret = lttng_syscall_filter_enable(event->chan, > +- desc->name); > ++ ret = lttng_syscall_filter_enable(event->chan, event); > + break; > + case LTTNG_KERNEL_KPROBE: > + case LTTNG_KERNEL_UPROBE: > +@@ -870,8 +895,7 @@ int _lttng_event_unregister(struct lttng_event *event) > + ret = 0; > + break; > + case LTTNG_KERNEL_SYSCALL: > +- ret = lttng_syscall_filter_disable(event->chan, > +- desc->name); > ++ ret = lttng_syscall_filter_disable(event->chan, event); > + break; > + case LTTNG_KERNEL_NOOP: > + ret = 0; > +@@ -1203,39 +1227,87 @@ int lttng_desc_match_enabler(const struct > lttng_event_desc *desc, > + struct lttng_enabler *enabler) > + { > + const char *desc_name, *enabler_name; > ++ bool compat = false, entry = false; > + > + enabler_name = enabler->event_param.name; > + switch (enabler->event_param.instrumentation) { > + case LTTNG_KERNEL_TRACEPOINT: > + desc_name = desc->name; > ++ switch (enabler->type) { > ++ case LTTNG_ENABLER_STAR_GLOB: > ++ return lttng_match_enabler_star_glob(desc_name, > enabler_name); > ++ case LTTNG_ENABLER_NAME: > ++ return lttng_match_enabler_name(desc_name, > enabler_name); > ++ default: > ++ return -EINVAL; > ++ } > + break; > + case LTTNG_KERNEL_SYSCALL: > + desc_name = desc->name; > +- if (!strncmp(desc_name, "compat_", strlen("compat_"))) > ++ if (!strncmp(desc_name, "compat_", strlen("compat_"))) { > + desc_name += strlen("compat_"); > ++ compat = true; > ++ } > + if (!strncmp(desc_name, "syscall_exit_", > + strlen("syscall_exit_"))) { > + desc_name += strlen("syscall_exit_"); > + } else if (!strncmp(desc_name, "syscall_entry_", > + strlen("syscall_entry_"))) { > + desc_name += strlen("syscall_entry_"); > ++ entry = true; > + } else { > + WARN_ON_ONCE(1); > + return -EINVAL; > + } > ++ switch (enabler->event_param.u.syscall.entryexit) { > ++ case LTTNG_KERNEL_SYSCALL_ENTRYEXIT: > ++ break; > ++ case LTTNG_KERNEL_SYSCALL_ENTRY: > ++ if (!entry) > ++ return 0; > ++ break; > ++ case LTTNG_KERNEL_SYSCALL_EXIT: > ++ if (entry) > ++ return 0; > ++ break; > ++ default: > ++ return -EINVAL; > ++ } > ++ switch (enabler->event_param.u.syscall.abi) { > ++ case LTTNG_KERNEL_SYSCALL_ABI_ALL: > ++ break; > ++ case LTTNG_KERNEL_SYSCALL_ABI_NATIVE: > ++ if (compat) > ++ return 0; > ++ break; > ++ case LTTNG_KERNEL_SYSCALL_ABI_COMPAT: > ++ if (!compat) > ++ return 0; > ++ break; > ++ default: > ++ return -EINVAL; > ++ } > ++ switch (enabler->event_param.u.syscall.match) { > ++ case LTTNG_SYSCALL_MATCH_NAME: > ++ switch (enabler->type) { > ++ case LTTNG_ENABLER_STAR_GLOB: > ++ return > lttng_match_enabler_star_glob(desc_name, enabler_name); > ++ case LTTNG_ENABLER_NAME: > ++ return lttng_match_enabler_name(desc_name, > enabler_name); > ++ default: > ++ return -EINVAL; > ++ } > ++ break; > ++ case LTTNG_SYSCALL_MATCH_NR: > ++ return -EINVAL; /* Not implemented. */ > ++ default: > ++ return -EINVAL; > ++ } > + break; > + default: > + WARN_ON_ONCE(1); > + return -EINVAL; > + } > +- switch (enabler->type) { > +- case LTTNG_ENABLER_STAR_GLOB: > +- return lttng_match_enabler_star_glob(desc_name, enabler_name); > +- case LTTNG_ENABLER_NAME: > +- return lttng_match_enabler_name(desc_name, enabler_name); > +- default: > +- return -EINVAL; > +- } > + } > + > + static > +@@ -1361,9 +1433,21 @@ void lttng_create_event_if_missing(struct > lttng_enabler *enabler) > + static > + int lttng_enabler_ref_events(struct lttng_enabler *enabler) > + { > +- struct lttng_session *session = enabler->chan->session; > ++ struct lttng_channel *chan = enabler->chan; > ++ struct lttng_session *session = chan->session; > + struct lttng_event *event; > + > ++ if (enabler->event_param.instrumentation == LTTNG_KERNEL_SYSCALL && > ++ enabler->event_param.u.syscall.entryexit == > LTTNG_KERNEL_SYSCALL_ENTRYEXIT && > ++ enabler->event_param.u.syscall.abi == > LTTNG_KERNEL_SYSCALL_ABI_ALL && > ++ enabler->event_param.u.syscall.match == > LTTNG_SYSCALL_MATCH_NAME && > ++ !strcmp(enabler->event_param.name, "*")) { > ++ if (enabler->enabled) > ++ WRITE_ONCE(chan->syscall_all, 1); > ++ else > ++ WRITE_ONCE(chan->syscall_all, 0); > ++ } > ++ > + /* First ensure that probe events are created for this enabler. */ > + lttng_create_event_if_missing(enabler); > + > +diff --git a/lttng-events.h b/lttng-events.h > +index a36a312..d4d9976 100644 > +--- a/lttng-events.h > ++++ b/lttng-events.h > +@@ -292,6 +292,16 @@ struct lttng_uprobe_handler { > + struct list_head node; > + }; > + > ++enum lttng_syscall_entryexit { > ++ LTTNG_SYSCALL_ENTRY, > ++ LTTNG_SYSCALL_EXIT, > ++}; > ++ > ++enum lttng_syscall_abi { > ++ LTTNG_SYSCALL_ABI_NATIVE, > ++ LTTNG_SYSCALL_ABI_COMPAT, > ++}; > ++ > + /* > + * lttng_event structure is referred to by the tracing fast path. It must be > + * kept small. > +@@ -318,6 +328,11 @@ struct lttng_event { > + struct inode *inode; > + struct list_head head; > + } uprobe; > ++ struct { > ++ char *syscall_name; > ++ enum lttng_syscall_entryexit entryexit; > ++ enum lttng_syscall_abi abi; > ++ } syscall; > + } u; > + struct list_head list; /* Event list in session */ > + unsigned int metadata_dumped:1; > +@@ -457,10 +472,10 @@ struct lttng_channel { > + struct lttng_syscall_filter *sc_filter; > + int header_type; /* 0: unset, 1: compact, 2: large */ > + enum channel_type channel_type; > ++ int syscall_all; > + unsigned int metadata_dumped:1, > + sys_enter_registered:1, > + sys_exit_registered:1, > +- syscall_all:1, > + tstate:1; /* Transient enable state */ > + }; > + > +@@ -653,10 +668,11 @@ void lttng_clock_unref(void); > + #if defined(CONFIG_HAVE_SYSCALL_TRACEPOINTS) > + int lttng_syscalls_register(struct lttng_channel *chan, void *filter); > + int lttng_syscalls_unregister(struct lttng_channel *chan); > ++int lttng_syscalls_destroy(struct lttng_channel *chan); > + int lttng_syscall_filter_enable(struct lttng_channel *chan, > +- const char *name); > ++ struct lttng_event *event); > + int lttng_syscall_filter_disable(struct lttng_channel *chan, > +- const char *name); > ++ struct lttng_event *event); > + long lttng_channel_syscall_mask(struct lttng_channel *channel, > + struct lttng_kernel_syscall_mask __user *usyscall_mask); > + #else > +@@ -670,14 +686,19 @@ static inline int lttng_syscalls_unregister(struct > lttng_channel *chan) > + return 0; > + } > + > ++static inline int lttng_syscalls_destroy(struct lttng_channel *chan) > ++{ > ++ return 0; > ++} > ++ > + static inline int lttng_syscall_filter_enable(struct lttng_channel *chan, > +- const char *name) > ++ struct lttng_event *event); > + { > + return -ENOSYS; > + } > + > + static inline int lttng_syscall_filter_disable(struct lttng_channel *chan, > +- const char *name) > ++ struct lttng_event *event); > + { > + return -ENOSYS; > + } > +diff --git a/lttng-syscalls.c b/lttng-syscalls.c > +index 97f1ba9..26cead6 100644 > +--- a/lttng-syscalls.c > ++++ b/lttng-syscalls.c > +@@ -367,8 +367,10 @@ const struct trace_syscall_entry compat_sc_exit_table[] > = { > + #undef CREATE_SYSCALL_TABLE > + > + struct lttng_syscall_filter { > +- DECLARE_BITMAP(sc, NR_syscalls); > +- DECLARE_BITMAP(sc_compat, NR_compat_syscalls); > ++ DECLARE_BITMAP(sc_entry, NR_syscalls); > ++ DECLARE_BITMAP(sc_exit, NR_syscalls); > ++ DECLARE_BITMAP(sc_compat_entry, NR_compat_syscalls); > ++ DECLARE_BITMAP(sc_compat_exit, NR_compat_syscalls); > + }; > + > + static void syscall_entry_unknown(struct lttng_event *event, > +@@ -391,29 +393,23 @@ void syscall_entry_probe(void *__data, struct pt_regs > *regs, long id) > + size_t table_len; > + > + if (unlikely(in_compat_syscall())) { > +- struct lttng_syscall_filter *filter; > +- > +- filter = lttng_rcu_dereference(chan->sc_filter); > +- if (filter) { > +- if (id < 0 || id >= NR_compat_syscalls > +- || !test_bit(id, filter->sc_compat)) { > +- /* System call filtered out. */ > +- return; > +- } > ++ struct lttng_syscall_filter *filter = chan->sc_filter; > ++ > ++ if (id < 0 || id >= NR_compat_syscalls > ++ || (!READ_ONCE(chan->syscall_all) && !test_bit(id, > filter->sc_compat_entry))) { > ++ /* System call filtered out. */ > ++ return; > + } > + table = compat_sc_table; > + table_len = ARRAY_SIZE(compat_sc_table); > + unknown_event = chan->sc_compat_unknown; > + } else { > +- struct lttng_syscall_filter *filter; > +- > +- filter = lttng_rcu_dereference(chan->sc_filter); > +- if (filter) { > +- if (id < 0 || id >= NR_syscalls > +- || !test_bit(id, filter->sc)) { > +- /* System call filtered out. */ > +- return; > +- } > ++ struct lttng_syscall_filter *filter = chan->sc_filter; > ++ > ++ if (id < 0 || id >= NR_syscalls > ++ || (!READ_ONCE(chan->syscall_all) && !test_bit(id, > filter->sc_entry))) { > ++ /* System call filtered out. */ > ++ return; > + } > + table = sc_table; > + table_len = ARRAY_SIZE(sc_table); > +@@ -545,29 +541,23 @@ void syscall_exit_probe(void *__data, struct pt_regs > *regs, long ret) > + > + id = syscall_get_nr(current, regs); > + if (unlikely(in_compat_syscall())) { > +- struct lttng_syscall_filter *filter; > +- > +- filter = lttng_rcu_dereference(chan->sc_filter); > +- if (filter) { > +- if (id < 0 || id >= NR_compat_syscalls > +- || !test_bit(id, filter->sc_compat)) { > +- /* System call filtered out. */ > +- return; > +- } > ++ struct lttng_syscall_filter *filter = chan->sc_filter; > ++ > ++ if (id < 0 || id >= NR_compat_syscalls > ++ || (!READ_ONCE(chan->syscall_all) && !test_bit(id, > filter->sc_compat_exit))) { > ++ /* System call filtered out. */ > ++ return; > + } > + table = compat_sc_exit_table; > + table_len = ARRAY_SIZE(compat_sc_exit_table); > + unknown_event = chan->compat_sc_exit_unknown; > + } else { > +- struct lttng_syscall_filter *filter; > +- > +- filter = lttng_rcu_dereference(chan->sc_filter); > +- if (filter) { > +- if (id < 0 || id >= NR_syscalls > +- || !test_bit(id, filter->sc)) { > +- /* System call filtered out. */ > +- return; > +- } > ++ struct lttng_syscall_filter *filter = chan->sc_filter; > ++ > ++ if (id < 0 || id >= NR_syscalls > ++ || (!READ_ONCE(chan->syscall_all) && !test_bit(id, > filter->sc_exit))) { > ++ /* System call filtered out. */ > ++ return; > + } > + table = sc_exit_table; > + table_len = ARRAY_SIZE(sc_exit_table); > +@@ -713,27 +703,23 @@ int fill_table(const struct trace_syscall_entry > *table, size_t table_len, > + memset(&ev, 0, sizeof(ev)); > + switch (type) { > + case SC_TYPE_ENTRY: > +- strncpy(ev.name, SYSCALL_ENTRY_STR, > +- LTTNG_KERNEL_SYM_NAME_LEN); > ++ ev.u.syscall.entryexit = LTTNG_KERNEL_SYSCALL_ENTRY; > ++ ev.u.syscall.abi = LTTNG_KERNEL_SYSCALL_ABI_NATIVE; > + break; > + case SC_TYPE_EXIT: > +- strncpy(ev.name, SYSCALL_EXIT_STR, > +- LTTNG_KERNEL_SYM_NAME_LEN); > ++ ev.u.syscall.entryexit = LTTNG_KERNEL_SYSCALL_EXIT; > ++ ev.u.syscall.abi = LTTNG_KERNEL_SYSCALL_ABI_NATIVE; > + break; > + case SC_TYPE_COMPAT_ENTRY: > +- strncpy(ev.name, COMPAT_SYSCALL_ENTRY_STR, > +- LTTNG_KERNEL_SYM_NAME_LEN); > ++ ev.u.syscall.entryexit = LTTNG_KERNEL_SYSCALL_ENTRY; > ++ ev.u.syscall.abi = LTTNG_KERNEL_SYSCALL_ABI_COMPAT; > + break; > + case SC_TYPE_COMPAT_EXIT: > +- strncpy(ev.name, COMPAT_SYSCALL_EXIT_STR, > +- LTTNG_KERNEL_SYM_NAME_LEN); > +- break; > +- default: > +- BUG_ON(1); > ++ ev.u.syscall.entryexit = LTTNG_KERNEL_SYSCALL_EXIT; > ++ ev.u.syscall.abi = LTTNG_KERNEL_SYSCALL_ABI_COMPAT; > + break; > + } > +- strncat(ev.name, desc->name, > +- LTTNG_KERNEL_SYM_NAME_LEN - strlen(ev.name) - 1); > ++ strncpy(ev.name, desc->name, LTTNG_KERNEL_SYM_NAME_LEN); > + ev.name[LTTNG_KERNEL_SYM_NAME_LEN - 1] = '\0'; > + ev.instrumentation = LTTNG_KERNEL_SYSCALL; > + chan_table[i] = _lttng_event_create(chan, &ev, filter, > +@@ -803,6 +789,8 @@ int lttng_syscalls_register(struct lttng_channel *chan, > void *filter) > + strncpy(ev.name, desc->name, LTTNG_KERNEL_SYM_NAME_LEN); > + ev.name[LTTNG_KERNEL_SYM_NAME_LEN - 1] = '\0'; > + ev.instrumentation = LTTNG_KERNEL_SYSCALL; > ++ ev.u.syscall.entryexit = LTTNG_KERNEL_SYSCALL_ENTRY; > ++ ev.u.syscall.abi = LTTNG_KERNEL_SYSCALL_ABI_NATIVE; > + chan->sc_unknown = _lttng_event_create(chan, &ev, filter, > + desc, > + ev.instrumentation); > +@@ -820,6 +808,8 @@ int lttng_syscalls_register(struct lttng_channel *chan, > void *filter) > + strncpy(ev.name, desc->name, LTTNG_KERNEL_SYM_NAME_LEN); > + ev.name[LTTNG_KERNEL_SYM_NAME_LEN - 1] = '\0'; > + ev.instrumentation = LTTNG_KERNEL_SYSCALL; > ++ ev.u.syscall.entryexit = LTTNG_KERNEL_SYSCALL_ENTRY; > ++ ev.u.syscall.abi = LTTNG_KERNEL_SYSCALL_ABI_COMPAT; > + chan->sc_compat_unknown = _lttng_event_create(chan, &ev, > filter, > + desc, > + ev.instrumentation); > +@@ -837,6 +827,8 @@ int lttng_syscalls_register(struct lttng_channel *chan, > void *filter) > + strncpy(ev.name, desc->name, LTTNG_KERNEL_SYM_NAME_LEN); > + ev.name[LTTNG_KERNEL_SYM_NAME_LEN - 1] = '\0'; > + ev.instrumentation = LTTNG_KERNEL_SYSCALL; > ++ ev.u.syscall.entryexit = LTTNG_KERNEL_SYSCALL_EXIT; > ++ ev.u.syscall.abi = LTTNG_KERNEL_SYSCALL_ABI_COMPAT; > + chan->compat_sc_exit_unknown = _lttng_event_create(chan, &ev, > + filter, desc, > + ev.instrumentation); > +@@ -854,6 +846,8 @@ int lttng_syscalls_register(struct lttng_channel *chan, > void *filter) > + strncpy(ev.name, desc->name, LTTNG_KERNEL_SYM_NAME_LEN); > + ev.name[LTTNG_KERNEL_SYM_NAME_LEN - 1] = '\0'; > + ev.instrumentation = LTTNG_KERNEL_SYSCALL; > ++ ev.u.syscall.entryexit = LTTNG_KERNEL_SYSCALL_EXIT; > ++ ev.u.syscall.abi = LTTNG_KERNEL_SYSCALL_ABI_NATIVE; > + chan->sc_exit_unknown = _lttng_event_create(chan, &ev, filter, > + desc, ev.instrumentation); > + WARN_ON_ONCE(!chan->sc_exit_unknown); > +@@ -883,6 +877,14 @@ int lttng_syscalls_register(struct lttng_channel *chan, > void *filter) > + if (ret) > + return ret; > + #endif > ++ > ++ if (!chan->sc_filter) { > ++ chan->sc_filter = kzalloc(sizeof(struct lttng_syscall_filter), > ++ GFP_KERNEL); > ++ if (!chan->sc_filter) > ++ return -ENOMEM; > ++ } > ++ > + if (!chan->sys_enter_registered) { > + ret = lttng_wrapper_tracepoint_probe_register("sys_enter", > + (void *) syscall_entry_probe, chan); > +@@ -930,7 +932,11 @@ int lttng_syscalls_unregister(struct lttng_channel > *chan) > + return ret; > + chan->sys_exit_registered = 0; > + } > +- /* lttng_event destroy will be performed by lttng_session_destroy() */ > ++ return 0; > ++} > ++ > ++int lttng_syscalls_destroy(struct lttng_channel *chan) > ++{ > + kfree(chan->sc_table); > + kfree(chan->sc_exit_table); > + #ifdef CONFIG_COMPAT > +@@ -993,136 +999,150 @@ uint32_t get_sc_tables_len(void) > + return ARRAY_SIZE(sc_table) + ARRAY_SIZE(compat_sc_table); > + } > + > +-int lttng_syscall_filter_enable(struct lttng_channel *chan, > +- const char *name) > ++static > ++const char *get_syscall_name(struct lttng_event *event) > + { > +- int syscall_nr, compat_syscall_nr, ret; > +- struct lttng_syscall_filter *filter; > ++ size_t prefix_len = 0; > + > +- WARN_ON_ONCE(!chan->sc_table); > ++ WARN_ON_ONCE(event->instrumentation != LTTNG_KERNEL_SYSCALL); > + > +- if (!name) { > +- /* Enable all system calls by removing filter */ > +- if (chan->sc_filter) { > +- filter = chan->sc_filter; > +- rcu_assign_pointer(chan->sc_filter, NULL); > +- synchronize_trace(); > +- kfree(filter); > ++ switch (event->u.syscall.entryexit) { > ++ case LTTNG_SYSCALL_ENTRY: > ++ switch (event->u.syscall.abi) { > ++ case LTTNG_SYSCALL_ABI_NATIVE: > ++ prefix_len = strlen(SYSCALL_ENTRY_STR); > ++ break; > ++ case LTTNG_SYSCALL_ABI_COMPAT: > ++ prefix_len = strlen(COMPAT_SYSCALL_ENTRY_STR); > ++ break; > + } > +- chan->syscall_all = 1; > +- return 0; > +- } > +- > +- if (!chan->sc_filter) { > +- if (chan->syscall_all) { > +- /* > +- * All syscalls are already enabled. > +- */ > +- return -EEXIST; > ++ break; > ++ case LTTNG_SYSCALL_EXIT: > ++ switch (event->u.syscall.abi) { > ++ case LTTNG_SYSCALL_ABI_NATIVE: > ++ prefix_len = strlen(SYSCALL_EXIT_STR); > ++ break; > ++ case LTTNG_SYSCALL_ABI_COMPAT: > ++ prefix_len = strlen(COMPAT_SYSCALL_EXIT_STR); > ++ break; > + } > +- filter = kzalloc(sizeof(struct lttng_syscall_filter), > +- GFP_KERNEL); > +- if (!filter) > +- return -ENOMEM; > +- } else { > +- filter = chan->sc_filter; > ++ break; > + } > +- syscall_nr = get_syscall_nr(name); > +- compat_syscall_nr = get_compat_syscall_nr(name); > +- if (syscall_nr < 0 && compat_syscall_nr < 0) { > +- ret = -ENOENT; > +- goto error; > ++ WARN_ON_ONCE(prefix_len == 0); > ++ return event->desc->name + prefix_len; > ++} > ++ > ++int lttng_syscall_filter_enable(struct lttng_channel *chan, > ++ struct lttng_event *event) > ++{ > ++ struct lttng_syscall_filter *filter = chan->sc_filter; > ++ const char *syscall_name; > ++ unsigned long *bitmap; > ++ int syscall_nr; > ++ > ++ WARN_ON_ONCE(!chan->sc_table); > ++ > ++ syscall_name = get_syscall_name(event); > ++ > ++ switch (event->u.syscall.abi) { > ++ case LTTNG_SYSCALL_ABI_NATIVE: > ++ syscall_nr = get_syscall_nr(syscall_name); > ++ break; > ++ case LTTNG_SYSCALL_ABI_COMPAT: > ++ syscall_nr = get_compat_syscall_nr(syscall_name); > ++ break; > ++ default: > ++ return -EINVAL; > + } > +- if (syscall_nr >= 0) { > +- if (test_bit(syscall_nr, filter->sc)) { > +- ret = -EEXIST; > +- goto error; > ++ if (syscall_nr < 0) > ++ return -ENOENT; > ++ > ++ > ++ switch (event->u.syscall.entryexit) { > ++ case LTTNG_SYSCALL_ENTRY: > ++ switch (event->u.syscall.abi) { > ++ case LTTNG_SYSCALL_ABI_NATIVE: > ++ bitmap = filter->sc_entry; > ++ break; > ++ case LTTNG_SYSCALL_ABI_COMPAT: > ++ bitmap = filter->sc_compat_entry; > ++ break; > + } > +- bitmap_set(filter->sc, syscall_nr, 1); > +- } > +- if (compat_syscall_nr >= 0) { > +- if (test_bit(compat_syscall_nr, filter->sc_compat)) { > +- ret = -EEXIST; > +- goto error; > ++ break; > ++ case LTTNG_SYSCALL_EXIT: > ++ switch (event->u.syscall.abi) { > ++ case LTTNG_SYSCALL_ABI_NATIVE: > ++ bitmap = filter->sc_exit; > ++ break; > ++ case LTTNG_SYSCALL_ABI_COMPAT: > ++ bitmap = filter->sc_compat_exit; > ++ break; > + } > +- bitmap_set(filter->sc_compat, compat_syscall_nr, 1); > ++ break; > ++ default: > ++ return -EINVAL; > + } > +- if (!chan->sc_filter) > +- rcu_assign_pointer(chan->sc_filter, filter); > ++ if (test_bit(syscall_nr, bitmap)) > ++ return -EEXIST; > ++ bitmap_set(bitmap, syscall_nr, 1); > + return 0; > +- > +-error: > +- if (!chan->sc_filter) > +- kfree(filter); > +- return ret; > + } > + > + int lttng_syscall_filter_disable(struct lttng_channel *chan, > +- const char *name) > ++ struct lttng_event *event) > + { > +- int syscall_nr, compat_syscall_nr, ret; > +- struct lttng_syscall_filter *filter; > ++ struct lttng_syscall_filter *filter = chan->sc_filter; > ++ const char *syscall_name; > ++ unsigned long *bitmap; > ++ int syscall_nr; > + > + WARN_ON_ONCE(!chan->sc_table); > + > +- if (!chan->sc_filter) { > +- if (!chan->syscall_all) > +- return -EEXIST; > +- filter = kzalloc(sizeof(struct lttng_syscall_filter), > +- GFP_KERNEL); > +- if (!filter) > +- return -ENOMEM; > +- /* Trace all system calls, then apply disable. */ > +- bitmap_set(filter->sc, 0, NR_syscalls); > +- bitmap_set(filter->sc_compat, 0, NR_compat_syscalls); > +- } else { > +- filter = chan->sc_filter; > ++ syscall_name = get_syscall_name(event); > ++ > ++ switch (event->u.syscall.abi) { > ++ case LTTNG_SYSCALL_ABI_NATIVE: > ++ syscall_nr = get_syscall_nr(syscall_name); > ++ break; > ++ case LTTNG_SYSCALL_ABI_COMPAT: > ++ syscall_nr = get_compat_syscall_nr(syscall_name); > ++ break; > ++ default: > ++ return -EINVAL; > + } > ++ if (syscall_nr < 0) > ++ return -ENOENT; > + > +- if (!name) { > +- /* Fail if all syscalls are already disabled. */ > +- if (bitmap_empty(filter->sc, NR_syscalls) > +- && bitmap_empty(filter->sc_compat, > +- NR_compat_syscalls)) { > +- ret = -EEXIST; > +- goto error; > +- } > + > +- /* Disable all system calls */ > +- bitmap_clear(filter->sc, 0, NR_syscalls); > +- bitmap_clear(filter->sc_compat, 0, NR_compat_syscalls); > +- goto apply_filter; > +- } > +- syscall_nr = get_syscall_nr(name); > +- compat_syscall_nr = get_compat_syscall_nr(name); > +- if (syscall_nr < 0 && compat_syscall_nr < 0) { > +- ret = -ENOENT; > +- goto error; > +- } > +- if (syscall_nr >= 0) { > +- if (!test_bit(syscall_nr, filter->sc)) { > +- ret = -EEXIST; > +- goto error; > ++ switch (event->u.syscall.entryexit) { > ++ case LTTNG_SYSCALL_ENTRY: > ++ switch (event->u.syscall.abi) { > ++ case LTTNG_SYSCALL_ABI_NATIVE: > ++ bitmap = filter->sc_entry; > ++ break; > ++ case LTTNG_SYSCALL_ABI_COMPAT: > ++ bitmap = filter->sc_compat_entry; > ++ break; > + } > +- bitmap_clear(filter->sc, syscall_nr, 1); > +- } > +- if (compat_syscall_nr >= 0) { > +- if (!test_bit(compat_syscall_nr, filter->sc_compat)) { > +- ret = -EEXIST; > +- goto error; > ++ break; > ++ case LTTNG_SYSCALL_EXIT: > ++ switch (event->u.syscall.abi) { > ++ case LTTNG_SYSCALL_ABI_NATIVE: > ++ bitmap = filter->sc_exit; > ++ break; > ++ case LTTNG_SYSCALL_ABI_COMPAT: > ++ bitmap = filter->sc_compat_exit; > ++ break; > + } > +- bitmap_clear(filter->sc_compat, compat_syscall_nr, 1); > ++ break; > ++ default: > ++ return -EINVAL; > + } > +-apply_filter: > +- if (!chan->sc_filter) > +- rcu_assign_pointer(chan->sc_filter, filter); > +- chan->syscall_all = 0; > +- return 0; > ++ if (!test_bit(syscall_nr, bitmap)) > ++ return -EEXIST; > ++ bitmap_clear(bitmap, syscall_nr, 1); > + > +-error: > +- if (!chan->sc_filter) > +- kfree(filter); > +- return ret; > ++ return 0; > + } > + > + static > +@@ -1236,6 +1256,9 @@ const struct file_operations lttng_syscall_list_fops = > { > + .release = seq_release, > + }; > + > ++/* > ++ * A syscall is enabled if it is traced for either entry or exit. > ++ */ > + long lttng_channel_syscall_mask(struct lttng_channel *channel, > + struct lttng_kernel_syscall_mask __user *usyscall_mask) > + { > +@@ -1262,8 +1285,9 @@ long lttng_channel_syscall_mask(struct lttng_channel > *channel, > + char state; > + > + if (channel->sc_table) { > +- if (filter) > +- state = test_bit(bit, filter->sc); > ++ if (!READ_ONCE(channel->syscall_all) && filter) > ++ state = test_bit(bit, filter->sc_entry) > ++ || test_bit(bit, filter->sc_exit); > + else > + state = 1; > + } else { > +@@ -1275,9 +1299,11 @@ long lttng_channel_syscall_mask(struct lttng_channel > *channel, > + char state; > + > + if (channel->compat_sc_table) { > +- if (filter) > ++ if (!READ_ONCE(channel->syscall_all) && filter) > + state = test_bit(bit - ARRAY_SIZE(sc_table), > +- filter->sc_compat); > ++ filter->sc_compat_entry) > ++ || test_bit(bit - > ARRAY_SIZE(sc_table), > ++ filter->sc_compat_exit); > + else > + state = 1; > + } else { > +-- > +2.19.1 > + > diff --git a/meta/recipes-kernel/lttng/lttng-modules_2.12.2.bb > b/meta/recipes-kernel/lttng/lttng-modules_2.12.2.bb > index 0e1a209ce8..e36b327a08 100644 > --- a/meta/recipes-kernel/lttng/lttng-modules_2.12.2.bb > +++ b/meta/recipes-kernel/lttng/lttng-modules_2.12.2.bb > @@ -11,6 +11,16 @@ include lttng-platforms.inc > SRC_URI = "https://lttng.org/files/${BPN}/${BPN}-${PV}.tar.bz2 \ > file://Makefile-Do-not-fail-if-CONFIG_TRACEPOINTS-is-not-en.patch > \ > file://BUILD_RUNTIME_BUG_ON-vs-gcc7.patch \ > + > file://0001-Kconfig-fix-dependency-issue-when-building-in-tree-w.patch \ > + > file://0002-fix-Move-mmutrace.h-into-the-mmu-sub-directory-v5.9.patch \ > + > file://0003-fix-KVM-x86-mmu-Make-kvm_mmu_page-definition-and-acc.patch \ > + > file://0004-fix-ext4-limit-the-length-of-per-inode-prealloc-list.patch \ > + > file://0005-fix-ext4-indicate-via-a-block-bitmap-read-is-prefetc.patch \ > + file://0006-fix-removal-of-smp_-read_barrier_depends-v5.9.patch \ > + file://0007-fix-writeback-Drop-I_DIRTY_TIME_EXPIRE-v5.9.patch \ > + > file://0008-fix-writeback-Fix-sync-livelock-due-to-b_dirty_time-.patch \ > + > file://0009-fix-version-ranges-for-ext4_discard_preallocations-a.patch \ > + file://0010-Fix-system-call-filter-table.patch \ > " > > SRC_URI[sha256sum] = > "df50bc3bd58679705714f17721acf619a8b0cedc694f8a97052aa5099626feca" > @@ -36,7 +46,7 @@ SRC_URI_class-devupstream = > "git://git.lttng.org/lttng-modules;branch=stable-2.1 > file://Makefile-Do-not-fail-if-CONFIG_TRACEPOINTS-is-not-en.patch > \ > file://BUILD_RUNTIME_BUG_ON-vs-gcc7.patch \ > " > -SRCREV_class-devupstream = "57ccbfa6a8a79c7b84394c2097efaf7935607aa5" > +SRCREV_class-devupstream = "ad594e3a953db1b0c3c059fde45b5a5494f6be78" > PV_class-devupstream = "2.12.2+git${SRCPV}" > S_class-devupstream = "${WORKDIR}/git" > SRCREV_FORMAT ?= "lttng_git" > -- > 2.19.1 > >
-=-=-=-=-=-=-=-=-=-=-=- Links: You receive all messages sent to this group. View/Reply Online (#142793): https://lists.openembedded.org/g/openembedded-core/message/142793 Mute This Topic: https://lists.openembedded.org/mt/76853782/21656 Group Owner: [email protected] Unsubscribe: https://lists.openembedded.org/g/openembedded-core/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
