Uprobes: Support SDT markers having reference count (semaphore)
Subsystems (5)
v8 cover Uprobes: Support SDT markers having reference count (semaphore)
v8 changes:
- Call delayed_uprobe_remove(uprobe, NULL) from put_uprobe() as well.
- Other minor optimizations.
v7: https://lkml.org/lkml/2018/7/31/20
Description:
Userspace Statically Defined Tracepoints[1] are dtrace style markers inside userspace applications. Applications like PostgreSQL, MySQL, Pthread, Perl, Python, Java, Ruby, Node.js, libvirt, QEMU, glib etc have these markers embedded in them. These markers are added by developer at important places in the code. Each marker source expands to a single nop instruction in the compiled code but there may be additional overhead for computing the marker arguments which expands to couple of instructions. In case the overhead is more, execution of it can be omitted by runtime if() condition when no one is tracing on the marker:
if (reference_counter > 0) {
Execute marker instructions;
}Default value of reference counter is 0. Tracer has to increment the reference counter before tracing on a marker and decrement it when done with the tracing.
Currently, perf tool has limited supports for SDT markers. I.e. it can not trace markers surrounded by reference counter. Also, it's not easy to add reference counter logic in userspace tool like perf, so basic idea for this patchset is to add reference counter logic in the a uprobe infrastructure. Ex,[2]
# cat tick.c
...
for (i = 0; i < 100; i++) {
DTRACE_PROBE1(tick, loop1, i);
if (TICK_LOOP2_ENABLED()) {
DTRACE_PROBE1(tick, loop2, i);
}
printf("hi: %d\n", i);
sleep(1);
}
... Here tick:loop1 is marker without reference counter where as tick:loop2 is surrounded by reference counter condition.
# perf buildid-cache --add /tmp/tick # perf probe sdt_tick:loop1 # perf probe sdt_tick:loop2
# perf stat -e sdt_tick:loop1,sdt_tick:loop2 -- /tmp/tick
hi: 0
hi: 1
hi: 2
^C
Performance counter stats for '/tmp/tick':
3 sdt_tick:loop1
0 sdt_tick:loop2
2.747086086 seconds time elapsedPerf failed to record data for tick:loop2. Same experiment with this patch series:
# ./perf buildid-cache --add /tmp/tick
# ./perf probe sdt_tick:loop2
# ./perf stat -e sdt_tick:loop2 /tmp/tick
hi: 0
hi: 1
hi: 2
^C
Performance counter stats for '/tmp/tick':
3 sdt_tick:loop2
2.561851452 seconds time elapsed[1] https://sourceware.org/systemtap/wiki/UserSpaceProbeImplementation
Ravi Bangoria (6):
Uprobes: Simplify uprobe_register() body
Uprobe: Additional argument arch_uprobe to uprobe_write_opcode()
Uprobes: Support SDT markers having reference count (semaphore) Uprobes/sdt: Prevent multiple reference counter for same uprobe trace_uprobe/sdt: Prevent multiple reference counter for same uprobe perf probe: Support SDT markers having reference counter (semaphore)
arch/arm/probes/uprobes/core.c | 2 +- arch/mips/kernel/uprobes.c | 2 +- include/linux/uprobes.h | 7 +- kernel/events/uprobes.c | 329 +++++++++++++++++++++++++++++++++++------ kernel/trace/trace.c | 2 +- kernel/trace/trace_uprobe.c | 75 +++++++++- tools/perf/util/probe-event.c | 39 ++++- tools/perf/util/probe-event.h | 1 + tools/perf/util/probe-file.c | 34 ++++- tools/perf/util/probe-file.h | 1 + tools/perf/util/symbol-elf.c | 46 ++++-- tools/perf/util/symbol.h | 7 + 12 files changed, 472 insertions(+), 73 deletions(-)
Signature · 1 line
2.14.4
PATCH v8 1/6 Uprobes: Simplify uprobe_register() body
Simplify uprobe_register() function body and let __uprobe_register() handle everything. Also move dependency functions around to fix build failures.
--- kernel/events/uprobes.c | 69 ++++++++++++++++++++++++++----------------------- 1 file changed, 36 insertions(+), 33 deletions(-)
kernel/events/uprobes.c +36 -33
@@ -840,13 +840,8 @@ register_for_each_vma(struct uprobe *uprobe, struct uprobe_consumer *new)return err; } -static int __uprobe_register (struct uprobe *uprobe,struct uprobe_consumer *uc) -{ -consumer_add (uprobe, uc); -return register_for_each_vma (uprobe, uc); -} - -static void __uprobe_unregister (struct uprobe *uprobe,struct uprobe_consumer *uc) +static void +__uprobe_unregister (struct uprobe *uprobe,struct uprobe_consumer *uc) {int err;
@@ -860,24 +855,46 @@ static void __uprobe_unregister(struct uprobe *uprobe, struct uprobe_consumer *u }/* -* uprobe_register - register a probe +* uprobe_unregister - unregister a already registered probe. +* @inode: the file in which the probe has to be removed. +* @offset: offset from the start of the file. +* @uc: identify which probe if multiple probes are colocated. +*/ +void uprobe_unregister (struct inode *inode,loff_t offset,struct uprobe_consumer *uc) +{ +struct uprobe *uprobe; + + uprobe =find_uprobe (inode, offset); +if (WARN_ON (!uprobe)) +return ; + +down_write (&uprobe->register_rwsem); +__uprobe_unregister (uprobe, uc); +up_write (&uprobe->register_rwsem); +put_uprobe (uprobe); +} +EXPORT_SYMBOL_GPL (uprobe_unregister); + +/* +* __uprobe_register - register a probe * @inode: the file in which the probe has to be placed. * @offset: offset from the start of the file. * @uc: information on howto handle the probe.. * -* Apart from the access refcount, uprobe_register() takes a creation +* Apart from the access refcount, __uprobe_register() takes a creation * refcount (thro alloc_uprobe) if and only if this @uprobe is getting * inserted into the rbtree (i.e first consumer for a @inode:@offset * tuple). Creation refcount stops uprobe_unregister from freeing the * @uprobe even before the register operation is complete. Creation * refcount is released when the last @uc for the @uprobe -* unregisters. Caller of uprobe_register() is required to keep @inode +* unregisters. Caller of __uprobe_register() is required to keep @inode * (and the containing mount) referenced. * * Return errno if it cannot successully install probes * else return 0 (success) */ -int uprobe_register (struct inode *inode,loff_t offset,struct uprobe_consumer *uc) +static int __uprobe_register (struct inode *inode,loff_t offset, +struct uprobe_consumer *uc) {struct uprobe *uprobe;int ret;
@@ -904,7 +921,8 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer *down_write (&uprobe->register_rwsem); ret = -EAGAIN ;if (likely (uprobe_is_active (uprobe))) { - ret =__uprobe_register (uprobe, uc); +consumer_add (uprobe, uc); + ret =register_for_each_vma (uprobe, uc);if (ret)__uprobe_unregister (uprobe, uc); }
@@ -915,6 +933,12 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer * goto retry;return ret; } + +int uprobe_register (struct inode *inode,loff_t offset, +struct uprobe_consumer *uc) +{ +return __uprobe_register (inode, offset, uc); +}EXPORT_SYMBOL_GPL (uprobe_register);/*
@@ -946,27 +970,6 @@ int uprobe_apply(struct inode *inode, loff_t offset,return ret; } -/* -* uprobe_unregister - unregister a already registered probe. -* @inode: the file in which the probe has to be removed. -* @offset: offset from the start of the file. -* @uc: identify which probe if multiple probes are colocated. -*/ -void uprobe_unregister (struct inode *inode,loff_t offset,struct uprobe_consumer *uc) -{ -struct uprobe *uprobe; - - uprobe =find_uprobe (inode, offset); -if (WARN_ON (!uprobe)) -return ; - -down_write (&uprobe->register_rwsem); -__uprobe_unregister (uprobe, uc); -up_write (&uprobe->register_rwsem); -put_uprobe (uprobe); -} -EXPORT_SYMBOL_GPL (uprobe_unregister); -static int unapply_uprobe (struct uprobe *uprobe,struct mm_struct *mm) {struct vm_area_struct *vma;
Signature · 1 line
2.14.4
Re: [PATCH v8 1/6] Uprobes: Simplify uprobe_register() body
Quoted 5 lines · Ravi Bangoria
Simplify
uprobe_register()function body and let__uprobe_register()handle everything. Also move dependency functions around to fix build failures.Signed-off-by Ravi Bangoria <[email protected]>
Quoted 129 lines
--- kernel/events/uprobes.c | 69 ++++++++++++++++++++++++++----------------------- 1 file changed, 36 insertions(+), 33 deletions(-)kernel/events/uprobes.c +36
-34@@ -840,13 +840,8 @@ register_for_each_vma(struct uprobe *uprobe, struct uprobe_consumer *new)return err; } -static int __uprobe_register (struct uprobe *uprobe,struct uprobe_consumer *uc) -{ -consumer_add (uprobe, uc); -return register_for_each_vma (uprobe, uc); -} - -static void __uprobe_unregister (struct uprobe *uprobe,struct uprobe_consumer *uc) +static void +__uprobe_unregister (struct uprobe *uprobe,struct uprobe_consumer *uc) {int err;@@ -860,24 +855,46 @@ static void __uprobe_unregister(struct uprobe *uprobe, struct uprobe_consumer *u }/* -* uprobe_register - register a probe +* uprobe_unregister - unregister a already registered probe. +* @inode: the file in which the probe has to be removed. +* @offset: offset from the start of the file. +* @uc: identify which probe if multiple probes are colocated. +*/ +void uprobe_unregister (struct inode *inode,loff_t offset,struct uprobe_consumer *uc) +{ +struct uprobe *uprobe; + + uprobe =find_uprobe (inode, offset); +if (WARN_ON (!uprobe)) +return ; + +down_write (&uprobe->register_rwsem); +__uprobe_unregister (uprobe, uc); +up_write (&uprobe->register_rwsem); +put_uprobe (uprobe); +} +EXPORT_SYMBOL_GPL (uprobe_unregister); + +/* +* __uprobe_register - register a probe * @inode: the file in which the probe has to be placed. * @offset: offset from the start of the file. * @uc: information on howto handle the probe.. * -* Apart from the access refcount, uprobe_register() takes a creation +* Apart from the access refcount, __uprobe_register() takes a creation * refcount (thro alloc_uprobe) if and only if this @uprobe is getting * inserted into the rbtree (i.e first consumer for a @inode:@offset * tuple). Creation refcount stops uprobe_unregister from freeing the * @uprobe even before the register operation is complete. Creation * refcount is released when the last @uc for the @uprobe -* unregisters. Caller of uprobe_register() is required to keep @inode +* unregisters. Caller of __uprobe_register() is required to keep @inode * (and the containing mount) referenced. * * Return errno if it cannot successully install probes * else return 0 (success) */ -int uprobe_register (struct inode *inode,loff_t offset,struct uprobe_consumer *uc) +static int __uprobe_register (struct inode *inode,loff_t offset, +struct uprobe_consumer *uc) {struct uprobe *uprobe;int ret;@@ -904,7 +921,8 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer *down_write (&uprobe->register_rwsem); ret = -EAGAIN ;if (likely (uprobe_is_active (uprobe))) { - ret =__uprobe_register (uprobe, uc); +consumer_add (uprobe, uc); + ret =register_for_each_vma (uprobe, uc);if (ret)__uprobe_unregister (uprobe, uc); }@@ -915,6 +933,12 @@ int uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer * goto retry;return ret; } + +int uprobe_register (struct inode *inode,loff_t offset, +struct uprobe_consumer *uc) +{ +return __uprobe_register (inode, offset, uc); +}EXPORT_SYMBOL_GPL (uprobe_register);/* @@ -946,27 +970,6 @@ int uprobe_apply(struct inode *inode, loff_t offset,return ret; } -/* -* uprobe_unregister - unregister a already registered probe. -* @inode: the file in which the probe has to be removed. -* @offset: offset from the start of the file. -* @uc: identify which probe if multiple probes are colocated. -*/ -void uprobe_unregister (struct inode *inode,loff_t offset,struct uprobe_consumer *uc) -{ -struct uprobe *uprobe; - - uprobe =find_uprobe (inode, offset); -if (WARN_ON (!uprobe)) -return ; - -down_write (&uprobe->register_rwsem); -__uprobe_unregister (uprobe, uc); -up_write (&uprobe->register_rwsem); -put_uprobe (uprobe); -} -EXPORT_SYMBOL_GPL (uprobe_unregister); -static int unapply_uprobe (struct uprobe *uprobe,struct mm_struct *mm) {struct vm_area_struct *vma; --2.14.4
Re: [PATCH v8 1/6] Uprobes: Simplify uprobe_register() body
In reply to Ravi BangoriaQuoted 4 lines · Ravi Bangoria
Simplify
uprobe_register()function body and let__uprobe_register()handle everything. Also move dependency functions around to fix build failures.
One nit:
s/to fix build/failures/to avoid build failures/
Quoted 2 lines
Signed-off-by Ravi Bangoria <[email protected]>---
Steven Rostedt
maintainerRe: [PATCH v8 1/6] Uprobes: Simplify uprobe_register() body
Quoted 9 lines · On Mon, 13 Aug 2018 01:56:12 -0700 Srikar Dronamraju…
Quoted 4 lines · Ravi Bangoria
Simplify
uprobe_register()function body and let__uprobe_register()handle everything. Also move dependency functions around to fix build failures.One nit:
s/to fix build/failures/to avoid build failures/
I pulled in this patch with the above update to the change log.
Quoted 8 lines
Quoted 5 lines
Signed-off-by Ravi Bangoria <[email protected]>---
Thanks, I added your ack and Song's Reviewed-by tags.
-- Steve
PATCH v8 2/6 Uprobe: Additional argument arch_uprobe to uprobe_write_opcode()
In reply to Ravi BangoriaAdd addition argument 'arch_uprobe' to uprobe_write_opcode(). We need this in later set of patches.
--- arch/arm/probes/uprobes/core.c | 2 +- arch/mips/kernel/uprobes.c | 2 +- include/linux/uprobes.h | 2 +- kernel/events/uprobes.c | 9 +++++---- 4 files changed, 8 insertions(+), 7 deletions(-)
arch/arm/probes/uprobes/core.c +1 -1
@@ -32,7 +32,7 @@ bool is_swbp_insn(uprobe_opcode_t *insn)int set_swbp (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr) { -return uprobe_write_opcode (mm, vaddr, +return uprobe_write_opcode (auprobe, mm, vaddr,__opcode_to_mem_arm (auprobe->bpinsn)); }
arch/mips/kernel/uprobes.c +1 -1
@@ -224,7 +224,7 @@ unsigned long arch_uretprobe_hijack_return_addr(int __weak set_swbp (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr) { -return uprobe_write_opcode (mm, vaddr,UPROBE_SWBP_INSN ); +return uprobe_write_opcode (auprobe, mm, vaddr,UPROBE_SWBP_INSN ); } void arch_uprobe_copy_ixol(struct page *page, unsigned long vaddr,
include/linux/uprobes.h +1 -1
@@ -121,7 +121,7 @@ extern bool is_swbp_insn(uprobe_opcode_t *insn);extern bool is_trap_insn (uprobe_opcode_t *insn);extern unsigned long uprobe_get_swbp_addr (struct pt_regs *regs);extern unsigned long uprobe_get_trap_addr (struct pt_regs *regs); -extern int uprobe_write_opcode (struct mm_struct *mm,unsigned long vaddr,uprobe_opcode_t ); +extern int uprobe_write_opcode (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr,uprobe_opcode_t );extern int uprobe_register (struct inode *inode,loff_t offset,struct uprobe_consumer *uc);extern int uprobe_apply (struct inode *inode,loff_t offset,struct uprobe_consumer *uc,bool );extern void uprobe_unregister (struct inode *inode,loff_t offset,struct uprobe_consumer *uc);
kernel/events/uprobes.c +5 -4
@@ -299,8 +299,8 @@ static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t* Called with mm->mmap_sem held for write. * Return 0 (success) or a negative errno. */ -int uprobe_write_opcode (struct mm_struct *mm,unsigned long vaddr, -uprobe_opcode_t opcode) +int uprobe_write_opcode (struct arch_uprobe *auprobe,struct mm_struct *mm, +unsigned long vaddr,uprobe_opcode_t opcode) {struct page *old_page, *new_page;struct vm_area_struct *vma;
@@ -351,7 +351,7 @@ int uprobe_write_opcode(struct mm_struct *mm, unsigned long vaddr,*/ int __weak set_swbp (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr) { -return uprobe_write_opcode (mm, vaddr,UPROBE_SWBP_INSN ); +return uprobe_write_opcode (auprobe, mm, vaddr,UPROBE_SWBP_INSN ); }/**
@@ -366,7 +366,8 @@ int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsignedint __weak set_orig_insn (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr) { -return uprobe_write_opcode (mm, vaddr, *(uprobe_opcode_t *)&auprobe->insn); +return uprobe_write_opcode (auprobe, mm, vaddr, + *(uprobe_opcode_t *)&auprobe->insn); }static struct uprobe *get_uprobe (struct uprobe *uprobe)
Signature · 1 line
2.14.4
Re: [PATCH v8 2/6] Uprobe: Additional argument arch_uprobe to uprobe_write_opcode()
Quoted 4 lines · Ravi Bangoria
Add addition argument '
arch_uprobe' touprobe_write_opcode(). We need this in later set of patches.Signed-off-by Ravi Bangoria <[email protected]>
Quoted 83 lines
--- arch/arm/probes/uprobes/core.c | 2 +- arch/mips/kernel/uprobes.c | 2 +- include/linux/uprobes.h | 2 +- kernel/events/uprobes.c | 9 +++++---- 4 files changed, 8 insertions(+), 7 deletions(-)arch/arm/probes/uprobes/core.c +1
-1@@ -32,7 +32,7 @@ bool is_swbp_insn(uprobe_opcode_t *insn)int set_swbp (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr) { -return uprobe_write_opcode (mm, vaddr, +return uprobe_write_opcode (auprobe, mm, vaddr,__opcode_to_mem_arm (auprobe->bpinsn)); }arch/mips/kernel/uprobes.c +1
-1@@ -224,7 +224,7 @@ unsigned long arch_uretprobe_hijack_return_addr(int __weak set_swbp (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr) { -return uprobe_write_opcode (mm, vaddr,UPROBE_SWBP_INSN ); +return uprobe_write_opcode (auprobe, mm, vaddr,UPROBE_SWBP_INSN ); } void arch_uprobe_copy_ixol(struct page *page, unsigned long vaddr,include/linux/uprobes.h +1
-1@@ -121,7 +121,7 @@ extern bool is_swbp_insn(uprobe_opcode_t *insn);extern bool is_trap_insn (uprobe_opcode_t *insn);extern unsigned long uprobe_get_swbp_addr (struct pt_regs *regs);extern unsigned long uprobe_get_trap_addr (struct pt_regs *regs); -extern int uprobe_write_opcode (struct mm_struct *mm,unsigned long vaddr,uprobe_opcode_t ); +extern int uprobe_write_opcode (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr,uprobe_opcode_t );extern int uprobe_register (struct inode *inode,loff_t offset,struct uprobe_consumer *uc);extern int uprobe_apply (struct inode *inode,loff_t offset,struct uprobe_consumer *uc,bool );extern void uprobe_unregister (struct inode *inode,loff_t offset,struct uprobe_consumer *uc);kernel/events/uprobes.c +5
-5@@ -299,8 +299,8 @@ static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t* Called with mm->mmap_sem held for write. * Return 0 (success) or a negative errno. */ -int uprobe_write_opcode (struct mm_struct *mm,unsigned long vaddr, -uprobe_opcode_t opcode) +int uprobe_write_opcode (struct arch_uprobe *auprobe,struct mm_struct *mm, +unsigned long vaddr,uprobe_opcode_t opcode) {struct page *old_page, *new_page;struct vm_area_struct *vma;@@ -351,7 +351,7 @@ int uprobe_write_opcode(struct mm_struct *mm, unsigned long vaddr,*/ int __weak set_swbp (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr) { -return uprobe_write_opcode (mm, vaddr,UPROBE_SWBP_INSN ); +return uprobe_write_opcode (auprobe, mm, vaddr,UPROBE_SWBP_INSN ); }/** @@ -366,7 +366,8 @@ int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsignedint __weak set_orig_insn (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr) { -return uprobe_write_opcode (mm, vaddr, *(uprobe_opcode_t *)&auprobe->insn); +return uprobe_write_opcode (auprobe, mm, vaddr, + *(uprobe_opcode_t *)&auprobe->insn); }static struct uprobe *get_uprobe (struct uprobe *uprobe) --2.14.4
Re: [PATCH v8 2/6] Uprobe: Additional argument arch_uprobe to uprobe_write_opcode()
In reply to Ravi BangoriaQuoted 4 lines · Ravi Bangoria
Add addition argument '
arch_uprobe' touprobe_write_opcode(). We need this in later set of patches.Signed-off-by Ravi Bangoria <[email protected]>
Looks good to me.
PATCH v8 3/6 Uprobes: Support SDT markers having reference count (semaphore)
In reply to Ravi BangoriaUserspace Statically Defined Tracepoints[1] are dtrace style markers inside userspace applications. Applications like PostgreSQL, MySQL, Pthread, Perl, Python, Java, Ruby, Node.js, libvirt, QEMU, glib etc have these markers embedded in them. These markers are added by developer at important places in the code. Each marker source expands to a single nop instruction in the compiled code but there may be additional overhead for computing the marker arguments which expands to couple of instructions. In case the overhead is more, execution of it can be omitted by runtime if() condition when no one is tracing on the marker:
if (reference_counter > 0) {
Execute marker instructions;
}Default value of reference counter is 0. Tracer has to increment the reference counter before tracing on a marker and decrement it when done with the tracing.
Implement the reference counter logic in core uprobe. User will be able to use it from trace_uprobe as well as from kernel module. New trace_uprobe definition with reference counter will now be:
<path>:<offset>[(ref_ctr_offset)]
where ref_ctr_offset is an optional field. For kernel module, new variant of uprobe_register() has been introduced:
uprobe_register_refctr(inode, offset, ref_ctr_offset, consumer)
No new variant for uprobe_unregister() because it's assumed to have only one reference counter for one uprobe.
[1] https://sourceware.org/systemtap/wiki/UserSpaceProbeImplementation
Note: 'reference counter' is called as 'semaphore' in original Dtrace (or Systemtap, bcc and even in ELF) documentation and code. But the term 'semaphore' is misleading in this context. This is just a counter used to hold number of tracers tracing on a marker. This is not really used for any synchronization. So we are referring it as 'reference counter' in kernel / perf code.
--- include/linux/uprobes.h | 5 + kernel/events/uprobes.c | 246 ++++++++++++++++++++++++++++++++++++++++++-- kernel/trace/trace.c | 2 +- kernel/trace/trace_uprobe.c | 38 ++++++- 4 files changed, 280 insertions(+), 11 deletions(-)
include/linux/uprobes.h +5 -0
@@ -123,6 +123,7 @@ extern unsigned long uprobe_get_swbp_addr(struct pt_regs *regs);extern unsigned long uprobe_get_trap_addr (struct pt_regs *regs);extern int uprobe_write_opcode (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr,uprobe_opcode_t );extern int uprobe_register (struct inode *inode,loff_t offset,struct uprobe_consumer *uc); +extern int uprobe_register_refctr (struct inode *inode,loff_t offset,loff_t ref_ctr_offset,struct uprobe_consumer *uc);extern int uprobe_apply (struct inode *inode,loff_t offset,struct uprobe_consumer *uc,bool );extern void uprobe_unregister (struct inode *inode,loff_t offset,struct uprobe_consumer *uc);extern int uprobe_mmap (struct vm_area_struct *vma);
@@ -160,6 +161,10 @@ uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer *uc) {return -ENOSYS ; } +static inline int uprobe_register_refctr (struct inode *inode,loff_t offset,loff_t ref_ctr_offset,struct uprobe_consumer *uc) +{ +return -ENOSYS ; +}static inline int uprobe_apply (struct inode *inode,loff_t offset,struct uprobe_consumer *uc,bool add) {
kernel/events/uprobes.c +239 -7
342 lines not on this page · raw message
kernel/trace/trace.c +1 -1
@@ -4620,7 +4620,7 @@ static const char readme_msg[] = "place (kretprobe): [<module>:]<symbol>[+<offset>]|<memaddr>\n"#endif #ifdef CONFIG_UPROBE_EVENTS - "\t place: <path>:<offset>\n" + " place (uprobe): <path>:<offset>[(ref_ctr_offset)]\n"#endif "\t args: <name>=fetcharg[:type]\n" "\t fetcharg: %<register>, @<address>, @<symbol>[+|-<offset>],\n"
kernel/trace/trace_uprobe.c +35 -3
@@ -59,6 +59,7 @@ struct trace_uprobe {struct inode *inode;char *filename;unsigned long offset; +unsigned long ref_ctr_offset;unsigned long nhit;struct trace_probe tp; };
@@ -364,10 +365,10 @@ static int register_trace_uprobe(struct trace_uprobe *tu)static int create_trace_uprobe (int argc,char **argv) {struct trace_uprobe *tu; -char *arg, *event, *group, *filename; +char *arg, *event, *group, *filename, *rctr, *rctr_end;char buf[MAX_EVENT_NAME_LEN ];struct path path; -unsigned long offset; +unsigned long offset, ref_ctr_offset;bool is_delete, is_return;int i, ret;
@@ -376,6 +377,7 @@ static int create_trace_uprobe(int argc, char **argv) is_return = false; event =NULL ; group =NULL ; + ref_ctr_offset =0 ;/* argc must be >= 1 */ if (argv[0 ][0 ] =='-' )
@@ -450,6 +452,26 @@ static int create_trace_uprobe(int argc, char **argv) goto fail_address_parse; } +/* Parse reference counter offset if specified. */ + rctr =strchr (arg,'(' ); +if (rctr) { + rctr_end =strchr (rctr,')' ); +if (rctr > rctr_end || *(rctr_end +1 ) !=0 ) { + ret = -EINVAL ; +pr_info ("Invalid reference counter offset.\n" ); + goto fail_address_parse; + } + + *rctr++ ='\0' ; + *rctr_end ='\0' ; + ret =kstrtoul (rctr,0 , &ref_ctr_offset); +if (ret) { +pr_info ("Invalid reference counter offset.\n" ); + goto fail_address_parse; + } + } + +/* Parse uprobe offset. */ ret =kstrtoul (arg,0 , &offset);if (ret) goto fail_address_parse;
@@ -484,6 +506,7 @@ static int create_trace_uprobe(int argc, char **argv) goto fail_address_parse; } tu->offset = offset; + tu->ref_ctr_offset = ref_ctr_offset; tu->path = path; tu->filename =kstrdup (filename,GFP_KERNEL );
@@ -602,6 +625,9 @@ static int probes_seq_show(struct seq_file *m, void *v)trace_event_name (&tu->tp.call), tu->filename, (int)(sizeof(void *) * 2), tu->offset); +if (tu->ref_ctr_offset) +seq_printf (m,"(0x%lx)" , tu->ref_ctr_offset); +for (i =0 ; i < tu->tp.nr_args; i++)seq_printf (m," %s=%s" , tu->tp.args[i].name, tu->tp.args[i].comm);
@@ -917,7 +943,13 @@ probe_event_enable(struct trace_uprobe *tu, struct trace_event_file *file, tu->consumer.filter = filter; tu->inode =d_real_inode (tu->path.dentry); - ret =uprobe_register (tu->inode, tu->offset, &tu->consumer); +if (tu->ref_ctr_offset) { + ret =uprobe_register_refctr (tu->inode, tu->offset, + tu->ref_ctr_offset, &tu->consumer); + }else { + ret =uprobe_register (tu->inode, tu->offset, &tu->consumer); + } +if (ret) goto err_buffer;
Signature · 1 line
2.14.4
Oleg Nesterov
maintainerI need to read this (hopefully final) version carefully. I'll try to do this before next Monday.
just one note,
Quoted 9 lines · Ravi Bangoria
kernel/events/uprobes.c +8
-0+static void delayed_uprobe_remove (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct list_head *pos, *q; +struct delayed_uprobe *du; + +if (!uprobe && !mm) +return ; +
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 348
Quoted 10 lines
kernel/events/uprobes.c +10
-0+list_for_each_safe (pos, q, &delayed_uprobe_list) { + du = list_entry(pos, struct delayed_uprobe, list); + +if (uprobe && mm && du->uprobe == uprobe && du->mm == mm) +delayed_uprobe_delete (du); +else if (!uprobe && du->mm == mm) +delayed_uprobe_delete (du); +else if (!mm && du->uprobe == uprobe) +delayed_uprobe_delete (du); + }
Sorry, I can't resist... this doesn't look very nice. How about
list_for_each_safe(pos, q, &delayed_uprobe_list) {
du = list_entry(pos, struct delayed_uprobe, list);
if (uprobe && du->uprobe != uprobe)
continue;
if (mm && du->mm != mm)
continue;
delayed_uprobe_delete();
}I won't insist, this is cosmetic after all, but please consider this change in case you will need to send v9.
Oleg.
Steven Rostedt
maintainerQuoted 3 lines · On Thu, 9 Aug 2018 16:38:28 +0200 Oleg Nesterov <[email protected]> wrote:
I need to read this (hopefully final) version carefully. I'll try to do this before next Monday.
Monday may be the opening of the merge window (more likely Sunday). Do you think this is good enough for pushing it in this late in the game, or do you think we should wait another cycle?
-- Steve
Quoted 7 lines · Steven Rostedt
Quoted 3 lines · On Thu, 9 Aug 2018 16:38:28 +0200 Oleg Nesterov <[email protected]> wrote:
I need to read this (hopefully final) version carefully. I'll try to do this before next Monday.
Quoted 5 lines
Monday may be the opening of the merge window (more likely Sunday). Do you think this is good enough for pushing it in this late in the game, or do you think we should wait another cycle?
-- Steve
We (Facebook) have use cases in production that would benefit from this work, so I would rather see this gets in sooner than later. After a brief review (I will more
careful review before Monday), I think this set is not likely to break existing uprobes
(those w/o ref_ctr). Therefore, I think it is safe to put it in this cycle.
Thanks,
Song
Steven Rostedt
maintainerQuoted 12 lines · On Fri, 10 Aug 2018 23:14:01 -0700 Song Liu <[email protected]> wrote:
Quoted 11 lines · Steven Rostedt
Quoted 3 lines · On Thu, 9 Aug 2018 16:38:28 +0200 Oleg Nesterov <[email protected]> wrote:
I need to read this (hopefully final) version carefully. I'll try to do this before next Monday.
Monday may be the opening of the merge window (more likely Sunday). Do you think this is good enough for pushing it in this late in the game, or do you think we should wait another cycle?
Quoted 9 lines
Quoted 1 line
-- Steve
We (Facebook) have use cases in production that would benefit from this work, so I would rather see this gets in sooner than later. After a brief review (I will more
careful review before Monday), I think this set is not likely to break existing uprobes
(those w/oref_ctr). Therefore, I think it is safe to put it in this cycle.
It's still going under review, and the merge window is now open. I'd prefer that this waits till the next merge window.
-- Steve
Steven Rostedt
maintainerQuoted 25 lines · On Mon, 13 Aug 2018 20:01:41 -0400 Steven Rostedt <[email protected]> wrote:
Quoted 12 lines · On Fri, 10 Aug 2018 23:14:01 -0700 Song Liu <[email protected]> wrote:
Quoted 11 lines · Steven Rostedt
Quoted 3 lines · On Thu, 9 Aug 2018 16:38:28 +0200 Oleg Nesterov <[email protected]> wrote:
I need to read this (hopefully final) version carefully. I'll try to do this before next Monday.
Monday may be the opening of the merge window (more likely Sunday). Do you think this is good enough for pushing it in this late in the game, or do you think we should wait another cycle?
Quoted 9 lines
Quoted 1 line
-- Steve
We (Facebook) have use cases in production that would benefit from this work, so I would rather see this gets in sooner than later. After a brief review (I will more
careful review before Monday), I think this set is not likely to break existing uprobes
(those w/oref_ctr). Therefore, I think it is safe to put it in this cycle.
Quoted 4 lines
It's still going under review, and the merge window is now open. I'd prefer that this waits till the next merge window.
But this said, patch 1&2 look simple enough to include this merge window. I'll pull them in and start testing.
-- Steve
Quoted 128 lines · Ravi Bangoria
Userspace Statically Defined Tracepoints[1] are dtrace style markers inside userspace applications. Applications like PostgreSQL, MySQL, Pthread, Perl, Python, Java, Ruby, Node.js, libvirt, QEMU, glib etc have these markers embedded in them. These markers are added by developer at important places in the code. Each marker source expands to a single nop instruction in the compiled code but there may be additional overhead for computing the marker arguments which expands to couple of instructions. In case the overhead is more, execution of it can be omitted by runtime
if()condition when no one is tracing on the marker:if (reference_counter > 0) { Execute marker instructions; }Default value of reference counter is 0. Tracer has to increment the reference counter before tracing on a marker and decrement it when done with the tracing.
Implement the reference counter logic in core uprobe. User will be able to use it from
trace_uprobeas well as from kernel module. Newtrace_uprobedefinition with reference counter will now be:<path>:<offset>[(ref_ctr_offset)]where
ref_ctr_offsetis an optional field. For kernel module, new variant ofuprobe_register()has been introduced:
uprobe_register_refctr(inode, offset, ref_ctr_offset, consumer)No new variant for
uprobe_unregister()because it's assumed to have only one reference counter for one uprobe.[1] https://sourceware.org/systemtap/wiki/UserSpaceProbeImplementation
Note: 'reference counter' is called as 'semaphore' in original Dtrace (or Systemtap, bcc and even in ELF) documentation and code. But the term 'semaphore' is misleading in this context. This is just a counter used to hold number of tracers tracing on a marker. This is not really used for any synchronization. So we are referring it as 'reference counter' in kernel / perf code.
Signed-off-by Ravi Bangoria <[email protected]>Reviewed-by Masami Hiramatsu <[email protected]>[Only trace_uprobe.c]--- include/linux/uprobes.h | 5 + kernel/events/uprobes.c | 246 ++++++++++++++++++++++++++++++++++++++++++-- kernel/trace/trace.c | 2 +- kernel/trace/trace_uprobe.c | 38 ++++++- 4 files changed, 280 insertions(+), 11 deletions(-)include/linux/uprobes.h +5
-0@@ -123,6 +123,7 @@ extern unsigned long uprobe_get_swbp_addr(struct pt_regs *regs);extern unsigned long uprobe_get_trap_addr (struct pt_regs *regs);extern int uprobe_write_opcode (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr,uprobe_opcode_t );extern int uprobe_register (struct inode *inode,loff_t offset,struct uprobe_consumer *uc); +extern int uprobe_register_refctr (struct inode *inode,loff_t offset,loff_t ref_ctr_offset,struct uprobe_consumer *uc);extern int uprobe_apply (struct inode *inode,loff_t offset,struct uprobe_consumer *uc,bool );extern void uprobe_unregister (struct inode *inode,loff_t offset,struct uprobe_consumer *uc);extern int uprobe_mmap (struct vm_area_struct *vma);@@ -160,6 +161,10 @@ uprobe_register(struct inode *inode, loff_t offset, struct uprobe_consumer *uc) {return -ENOSYS ; } +static inline int uprobe_register_refctr (struct inode *inode,loff_t offset,loff_t ref_ctr_offset,struct uprobe_consumer *uc) +{ +return -ENOSYS ; +}static inline int uprobe_apply (struct inode *inode,loff_t offset,struct uprobe_consumer *uc,bool add) {kernel/events/uprobes.c +32
-0@@ -73,6 +73,7 @@ struct uprobe {struct uprobe_consumer *consumers;struct inode *inode;/* Also hold a ref to inode */ loff_t offset; +loff_t ref_ctr_offset;unsigned long flags;/* @@ -88,6 +89,15 @@ struct uprobe {struct arch_uprobe arch; }; +struct delayed_uprobe { +struct list_head list; +struct uprobe *uprobe; +struct mm_struct *mm; +}; + +static DEFINE_MUTEX (delayed_uprobe_lock); +static LIST_HEAD (delayed_uprobe_list); +/* * Execute out of line area: anonymous executable mapping installed * by the probed task to execute the copy of the original instruction @@ -282,6 +292,157 @@ static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_treturn 1 ; } +static struct delayed_uprobe * +delayed_uprobe_check (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct delayed_uprobe *du; + +list_for_each_entry (du, &delayed_uprobe_list, list) +if (du->uprobe == uprobe && du->mm == mm) +return du; +return NULL ; +} + +static int delayed_uprobe_add (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct delayed_uprobe *du; + +if (delayed_uprobe_check (uprobe, mm)) +return 0 ; + + du =kzalloc (sizeof (*du),GFP_KERNEL ); +if (!du) +return -ENOMEM ; +
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 326
Quoted 10 lines
kernel/events/uprobes.c +10
-0+ du->uprobe = uprobe; + du->mm = mm; +list_add (&du->list, &delayed_uprobe_list); +return 0 ; +} + +static void delayed_uprobe_delete (struct delayed_uprobe *du) +{ +if (!du) +return ;
Do we really need this check?
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 337
Quoted 11 lines
kernel/events/uprobes.c +11
-0+list_del (&du->list); +kfree (du); +} + +static void delayed_uprobe_remove (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct list_head *pos, *q; +struct delayed_uprobe *du; + +if (!uprobe && !mm) +return ;
And do we really need this check?
Quoted 35 lines
kernel/events/uprobes.c +35
-0+ +list_for_each_safe (pos, q, &delayed_uprobe_list) { + du = list_entry(pos, struct delayed_uprobe, list); + +if (uprobe && mm && du->uprobe == uprobe && du->mm == mm) +delayed_uprobe_delete (du); +else if (!uprobe && du->mm == mm) +delayed_uprobe_delete (du); +else if (!mm && du->uprobe == uprobe) +delayed_uprobe_delete (du); + } +} + +static bool valid_ref_ctr_vma (struct uprobe *uprobe, +struct vm_area_struct *vma) +{ +unsigned long vaddr =offset_to_vaddr (vma, uprobe->ref_ctr_offset); + +return uprobe->ref_ctr_offset && + vma->vm_file && +file_inode (vma->vm_file) == uprobe->inode && + (vma->vm_flags & (VM_WRITE |VM_SHARED )) ==VM_WRITE && + vma->vm_start <= vaddr && + vma->vm_end > vaddr; +} + +static struct vm_area_struct * +find_ref_ctr_vma (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct vm_area_struct *tmp; + +for (tmp = mm->mmap; tmp; tmp = tmp->vm_next) +if (valid_ref_ctr_vma (uprobe, tmp)) +return tmp; +
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 382
Quoted 10 lines
kernel/events/uprobes.c +10
-0+return NULL ; +} + +static int +__update_ref_ctr (struct mm_struct *mm,unsigned long vaddr,short d) +{ +void *kaddr; +struct page *page; +struct vm_area_struct *vma; +int ret =0 ;
It is not necessary to initialize ret here.
Quoted 5 lines
kernel/events/uprobes.c +5
-0+short *ptr; + +if (vaddr ==0 || d ==0 ) +return -EINVAL ; +
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 397
Quoted 10 lines
kernel/events/uprobes.c +10
-0+ ret =get_user_pages_remote (NULL , mm, vaddr,1 , +FOLL_WRITE , &page, &vma,NULL ); +if (unlikely (ret <=0 )) { +/* +* We are asking for 1 page. If get_user_pages_remote() fails, +* it may return 0, in that case we have to return error. +*/ + ret = (ret ==0 ) ? -EBUSY : ret; +pr_warn ("Failed to %s ref_ctr. (%d)\n" , + d >0 ?"increment" :"decrement" , ret);
This warning is not really useful. Seems this function has little information about which uprobe is failing here. Maybe we only need warning in the caller (or caller of caller).
Quoted 13 lines
kernel/events/uprobes.c +13
-0+return ret; + } + + kaddr =kmap_atomic (page); + ptr = kaddr + (vaddr & ~PAGE_MASK ); + +if (unlikely (*ptr + d <0 )) { +pr_warn ("ref_ctr going negative. vaddr: 0x%lx, " +"curr val: %d, delta: %d\n" , vaddr, *ptr, d); + ret = -EINVAL ; + goto out; + } +
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 420
Quoted 10 lines
kernel/events/uprobes.c +10
-0+ *ptr += d; + ret =0 ; +out: +kunmap_atomic (kaddr); +put_page (page); +return ret; +} + +static int update_ref_ctr (struct uprobe *uprobe,struct mm_struct *mm, +bool is_register)
What's the reason of bool is_register here vs. short d in __update_ref_ctr()? Can we use short for both?
Quoted 4 lines
kernel/events/uprobes.c +4
-0+{ +struct vm_area_struct *rc_vma; +unsigned long rc_vaddr; +int ret =0 ;
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 434
Quoted 10 lines
kernel/events/uprobes.c +10
-0+ + rc_vma =find_ref_ctr_vma (uprobe, mm); + +if (rc_vma) { + rc_vaddr =offset_to_vaddr (rc_vma, uprobe->ref_ctr_offset); + ret =__update_ref_ctr (mm, rc_vaddr, is_register ?1 :-1 ); + +if (is_register) +return ret; + }
Mixing __update_ref_ctr() here and delayed_uprobe_add() in the same function is a little confusing (at least for me). How about we always use delayed uprobe for uprobe_mmap() and use non-delayed in other case(s)?
Quoted 130 lines
kernel/events/uprobes.c +51
-6+ +mutex_lock (&delayed_uprobe_lock); +if (is_register) + ret =delayed_uprobe_add (uprobe, mm); +else +delayed_uprobe_remove (uprobe, mm); +mutex_unlock (&delayed_uprobe_lock); + +return ret; +} +/* * NOTE: * Expect the breakpoint instruction to be the smallest size instruction for @@ -302,9 +463,13 @@ static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_tint uprobe_write_opcode (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr,uprobe_opcode_t opcode) { +struct uprobe *uprobe;struct page *old_page, *new_page;struct vm_area_struct *vma; -int ret; +int ret, is_register, ref_ctr_updated =0 ; + + is_register =is_swbp_insn (&opcode); + uprobe = container_of(auprobe, struct uprobe, arch); retry:/* Read the page with vaddr into memory */ @@ -317,6 +482,15 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,if (ret <=0 ) goto put_old; +/* We are going to replace instruction, update ref_ctr. */ +if (!ref_ctr_updated && uprobe->ref_ctr_offset) { + ret =update_ref_ctr (uprobe, mm, is_register); +if (ret) + goto put_old; + + ref_ctr_updated =1 ; + } + ret =anon_vma_prepare (vma);if (ret) goto put_old;@@ -337,6 +511,11 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,if (unlikely (ret == -EAGAIN )) goto retry; + +/* Revert back reference counter if instruction update failed. */ +if (ret && is_register && ref_ctr_updated) +update_ref_ctr (uprobe, mm, false); +return ret; }@@ -378,8 +557,15 @@ static struct uprobe *get_uprobe(struct uprobe *uprobe)static void put_uprobe (struct uprobe *uprobe) { -if (atomic_dec_and_test (&uprobe->ref)) +if (atomic_dec_and_test (&uprobe->ref)) { +/* +* If application munmap(exec_vma) before uprobe_unregister() +* gets called, we don't get a chance to remove uprobe from +* delayed_uprobe_list in remove_breakpoint(). Do it here. +*/ +delayed_uprobe_remove (uprobe,NULL );kfree (uprobe); + } }static int match_uprobe (struct uprobe *l,struct uprobe *r)@@ -484,7 +670,8 @@ static struct uprobe *insert_uprobe(struct uprobe *uprobe)return u; } -static struct uprobe *alloc_uprobe (struct inode *inode,loff_t offset) +static struct uprobe *alloc_uprobe (struct inode *inode,loff_t offset, +loff_t ref_ctr_offset) {struct uprobe *uprobe, *cur_uprobe;@@ -494,6 +681,7 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset) uprobe->inode = inode; uprobe->offset = offset; + uprobe->ref_ctr_offset = ref_ctr_offset;init_rwsem (&uprobe->register_rwsem);init_rwsem (&uprobe->consumer_rwsem);@@ -895,7 +1083,7 @@ EXPORT_SYMBOL_GPL(uprobe_unregister);* else return 0 (success) */ static int __uprobe_register (struct inode *inode,loff_t offset, -struct uprobe_consumer *uc) +loff_t ref_ctr_offset,struct uprobe_consumer *uc) {struct uprobe *uprobe;int ret;@@ -912,7 +1100,7 @@ static int __uprobe_register(struct inode *inode, loff_t offset,return -EINVAL ; retry: - uprobe =alloc_uprobe (inode, offset); + uprobe =alloc_uprobe (inode, offset, ref_ctr_offset);if (!uprobe)return -ENOMEM ;/* @@ -938,10 +1126,17 @@ static int __uprobe_register(struct inode *inode, loff_t offset,int uprobe_register (struct inode *inode,loff_t offset,struct uprobe_consumer *uc) { -return __uprobe_register (inode, offset, uc); +return __uprobe_register (inode, offset,0 , uc); }EXPORT_SYMBOL_GPL (uprobe_register); +int uprobe_register_refctr (struct inode *inode,loff_t offset, +loff_t ref_ctr_offset,struct uprobe_consumer *uc) +{ +return __uprobe_register (inode, offset, ref_ctr_offset, uc); +} +EXPORT_SYMBOL_GPL (uprobe_register_refctr); +/* * uprobe_apply - unregister a already registered probe. * @inode: the file in which the probe has to be removed.
kernel/events/uprobes.c static void build_probe_list(struct inode *inode, · line 1258
Quoted 4 lines
kernel/events/uprobes.c +1
-0@@ -1060,6 +1255,31 @@ static void build_probe_list(struct inode *inode,spin_unlock (&uprobes_treelock); } +static int delayed_uprobe_install (struct vm_area_struct *vma)
This function name is confusing. How about we call it delayed_ref_ctr_incr() or something similar? Also, we should add comments to highlight this is vma is not the vma containing the uprobe, but the vma containing the ref_ctr.
Quoted 8 lines
kernel/events/uprobes.c +8
-0+{ +struct list_head *pos, *q; +struct delayed_uprobe *du; +unsigned long vaddr; +int ret =0 , err =0 ; + +mutex_lock (&delayed_uprobe_lock); +list_for_each_safe (pos, q, &delayed_uprobe_list) {
kernel/events/uprobes.c static void build_probe_list(struct inode *inode, · line 1276
Quoted 10 lines
kernel/events/uprobes.c +10
-0+ du = list_entry(pos, struct delayed_uprobe, list); + +if (!valid_ref_ctr_vma (du->uprobe, vma)) +continue ; + + vaddr =offset_to_vaddr (vma, du->uprobe->ref_ctr_offset); + ret =__update_ref_ctr (vma->vm_mm, vaddr,1 ); +/* Record an error and continue. */ +if (ret && !err) + err = ret;
I think this is a good place (when ret != 0) to call pr_warn(). I guess we can print which mm get error for which uprobe (inode+offset).
Quoted 146 lines
kernel/events/uprobes.c +19
-1+delayed_uprobe_delete (du); + } +mutex_unlock (&delayed_uprobe_lock); +return err; +} +/* * Called from mmap_region/vma_adjust with mm->mmap_sem acquired. * @@ -1072,7 +1292,15 @@ int uprobe_mmap(struct vm_area_struct *vma)struct uprobe *uprobe, *u;struct inode *inode; -if (no_uprobe_events () || !valid_vma (vma, true)) +if (no_uprobe_events ()) +return 0 ; + +if (vma->vm_file && + (vma->vm_flags & (VM_WRITE |VM_SHARED )) ==VM_WRITE && +test_bit (MMF_HAS_UPROBES , &vma->vm_mm->flags)) +delayed_uprobe_install (vma); + +if (!valid_vma (vma, true))return 0 ; inode =file_inode (vma->vm_file);@@ -1246,6 +1474,10 @@ void uprobe_clear_state(struct mm_struct *mm) {struct xol_area *area = mm->uprobes_state.xol_area; +mutex_lock (&delayed_uprobe_lock); +delayed_uprobe_remove (NULL , mm); +mutex_unlock (&delayed_uprobe_lock); +if (!area)return ;kernel/trace/trace.c +1
-1@@ -4620,7 +4620,7 @@ static const char readme_msg[] = "place (kretprobe): [<module>:]<symbol>[+<offset>]|<memaddr>\n"#endif #ifdef CONFIG_UPROBE_EVENTS - "\t place: <path>:<offset>\n" + " place (uprobe): <path>:<offset>[(ref_ctr_offset)]\n"#endif "\t args: <name>=fetcharg[:type]\n" "\t fetcharg: %<register>, @<address>, @<symbol>[+|-<offset>],\n" kernel/trace/trace_uprobe.c +35
-4@@ -59,6 +59,7 @@ struct trace_uprobe {struct inode *inode;char *filename;unsigned long offset; +unsigned long ref_ctr_offset;unsigned long nhit;struct trace_probe tp; };@@ -364,10 +365,10 @@ static int register_trace_uprobe(struct trace_uprobe *tu)static int create_trace_uprobe (int argc,char **argv) {struct trace_uprobe *tu; -char *arg, *event, *group, *filename; +char *arg, *event, *group, *filename, *rctr, *rctr_end;char buf[MAX_EVENT_NAME_LEN ];struct path path; -unsigned long offset; +unsigned long offset, ref_ctr_offset;bool is_delete, is_return;int i, ret;@@ -376,6 +377,7 @@ static int create_trace_uprobe(int argc, char **argv) is_return = false; event =NULL ; group =NULL ; + ref_ctr_offset =0 ;/* argc must be >= 1 */ if (argv[0 ][0 ] =='-' )@@ -450,6 +452,26 @@ static int create_trace_uprobe(int argc, char **argv) goto fail_address_parse; } +/* Parse reference counter offset if specified. */ + rctr =strchr (arg,'(' ); +if (rctr) { + rctr_end =strchr (rctr,')' ); +if (rctr > rctr_end || *(rctr_end +1 ) !=0 ) { + ret = -EINVAL ; +pr_info ("Invalid reference counter offset.\n" ); + goto fail_address_parse; + } + + *rctr++ ='\0' ; + *rctr_end ='\0' ; + ret =kstrtoul (rctr,0 , &ref_ctr_offset); +if (ret) { +pr_info ("Invalid reference counter offset.\n" ); + goto fail_address_parse; + } + } + +/* Parse uprobe offset. */ ret =kstrtoul (arg,0 , &offset);if (ret) goto fail_address_parse;@@ -484,6 +506,7 @@ static int create_trace_uprobe(int argc, char **argv) goto fail_address_parse; } tu->offset = offset; + tu->ref_ctr_offset = ref_ctr_offset; tu->path = path; tu->filename =kstrdup (filename,GFP_KERNEL );@@ -602,6 +625,9 @@ static int probes_seq_show(struct seq_file *m, void *v)trace_event_name (&tu->tp.call), tu->filename, (int)(sizeof(void *) * 2), tu->offset); +if (tu->ref_ctr_offset) +seq_printf (m,"(0x%lx)" , tu->ref_ctr_offset); +for (i =0 ; i < tu->tp.nr_args; i++)seq_printf (m," %s=%s" , tu->tp.args[i].name, tu->tp.args[i].comm);@@ -917,7 +943,13 @@ probe_event_enable(struct trace_uprobe *tu, struct trace_event_file *file, tu->consumer.filter = filter; tu->inode =d_real_inode (tu->path.dentry); - ret =uprobe_register (tu->inode, tu->offset, &tu->consumer); +if (tu->ref_ctr_offset) { + ret =uprobe_register_refctr (tu->inode, tu->offset, + tu->ref_ctr_offset, &tu->consumer); + }else { + ret =uprobe_register (tu->inode, tu->offset, &tu->consumer); + } +if (ret) goto err_buffer; --2.14.4
Steven Rostedt
maintainerQuoted 6 lines · On Sat, 11 Aug 2018 00:57:12 -0700 Song Liu <[email protected]> wrote:
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 326
Quoted 5 lines
kernel/events/uprobes.c +5
-0+ +static void delayed_uprobe_delete (struct delayed_uprobe *du) +{ +if (!du) +return ;Do we really need this check?
I'd suggest we keep it. It's not a fast path, and operations like this should really check for NULL.
Quoted 13 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 337
Quoted 12 lines
kernel/events/uprobes.c +11
-0+list_del (&du->list); +kfree (du); +} + +static void delayed_uprobe_remove (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct list_head *pos, *q; +struct delayed_uprobe *du; + +if (!uprobe && !mm) +return ;And do we really need this check?
Same here, as it's not a fast path, and it prevents kernel oops if a NULL is passed in.
Quoted 39 lines
Quoted 39 lines
kernel/events/uprobes.c +38
-0+ +list_for_each_safe (pos, q, &delayed_uprobe_list) { + du = list_entry(pos, struct delayed_uprobe, list); + +if (uprobe && mm && du->uprobe == uprobe && du->mm == mm) +delayed_uprobe_delete (du); +else if (!uprobe && du->mm == mm) +delayed_uprobe_delete (du); +else if (!mm && du->uprobe == uprobe) +delayed_uprobe_delete (du); + } +} + +static bool valid_ref_ctr_vma (struct uprobe *uprobe, +struct vm_area_struct *vma) +{ +unsigned long vaddr =offset_to_vaddr (vma, uprobe->ref_ctr_offset); + +return uprobe->ref_ctr_offset && + vma->vm_file && +file_inode (vma->vm_file) == uprobe->inode && + (vma->vm_flags & (VM_WRITE |VM_SHARED )) ==VM_WRITE && + vma->vm_start <= vaddr && + vma->vm_end > vaddr; +} + +static struct vm_area_struct * +find_ref_ctr_vma (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct vm_area_struct *tmp; + +for (tmp = mm->mmap; tmp; tmp = tmp->vm_next) +if (valid_ref_ctr_vma (uprobe, tmp)) +return tmp; + +return NULL ; +} +
Quoted 8 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 382
Quoted 7 lines
kernel/events/uprobes.c +7
-0+static int +__update_ref_ctr (struct mm_struct *mm,unsigned long vaddr,short d) +{ +void *kaddr; +struct page *page; +struct vm_area_struct *vma; +int ret =0 ;It is not necessary to initialize ret here.
Agreed.
Quoted 11 lines
Quoted 11 lines
kernel/events/uprobes.c +10
-0+short *ptr; + +if (vaddr ==0 || d ==0 ) +return -EINVAL ; + + ret =get_user_pages_remote (NULL , mm, vaddr,1 , +FOLL_WRITE , &page, &vma,NULL ); +if (unlikely (ret <=0 )) { +/* +* We are asking for 1 page. If get_user_pages_remote() fails,
Quoted 8 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 397
Quoted 5 lines
kernel/events/uprobes.c +5
-0+* it may return 0, in that case we have to return error. +*/ + ret = (ret ==0 ) ? -EBUSY : ret; +pr_warn ("Failed to %s ref_ctr. (%d)\n" , + d >0 ?"increment" :"decrement" , ret);This warning is not really useful. Seems this function has little information about which uprobe is failing here. Maybe we only need warning in the caller (or caller of caller).
I'm fine with or without this.
Quoted 18 lines
Quoted 18 lines
kernel/events/uprobes.c +17
-0+return ret; + } + + kaddr =kmap_atomic (page); + ptr = kaddr + (vaddr & ~PAGE_MASK ); + +if (unlikely (*ptr + d <0 )) { +pr_warn ("ref_ctr going negative. vaddr: 0x%lx, " +"curr val: %d, delta: %d\n" , vaddr, *ptr, d); + ret = -EINVAL ; + goto out; + } + + *ptr += d; + ret =0 ; +out: +kunmap_atomic (kaddr);
Quoted 8 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 420
Quoted 6 lines
kernel/events/uprobes.c +6
-0+put_page (page); +return ret; +} + +static int update_ref_ctr (struct uprobe *uprobe,struct mm_struct *mm, +bool is_register)What's the reason of bool
is_registerhere vs. short d in__update_ref_ctr()? Can we use short for both?
bool has more control, but I agree that this function is a bit confusing.
-- Steve
Quoted 324 lines
Quoted 5 lines
kernel/events/uprobes.c +4
-0+{ +struct vm_area_struct *rc_vma; +unsigned long rc_vaddr; +int ret =0 ;kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 434
Quoted 10 lines
kernel/events/uprobes.c +10
-0+ + rc_vma =find_ref_ctr_vma (uprobe, mm); + +if (rc_vma) { + rc_vaddr =offset_to_vaddr (rc_vma, uprobe->ref_ctr_offset); + ret =__update_ref_ctr (mm, rc_vaddr, is_register ?1 :-1 ); + +if (is_register) +return ret; + }Mixing
__update_ref_ctr()here anddelayed_uprobe_add()in the same function is a little confusing (at least for me). How about we always use delayed uprobe foruprobe_mmap()and use non-delayed in other case(s)?Quoted 130 lines
kernel/events/uprobes.c +51
-6+ +mutex_lock (&delayed_uprobe_lock); +if (is_register) + ret =delayed_uprobe_add (uprobe, mm); +else +delayed_uprobe_remove (uprobe, mm); +mutex_unlock (&delayed_uprobe_lock); + +return ret; +} +/* * NOTE: * Expect the breakpoint instruction to be the smallest size instruction for @@ -302,9 +463,13 @@ static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_tint uprobe_write_opcode (struct arch_uprobe *auprobe,struct mm_struct *mm,unsigned long vaddr,uprobe_opcode_t opcode) { +struct uprobe *uprobe;struct page *old_page, *new_page;struct vm_area_struct *vma; -int ret; +int ret, is_register, ref_ctr_updated =0 ; + + is_register =is_swbp_insn (&opcode); + uprobe = container_of(auprobe, struct uprobe, arch); retry:/* Read the page with vaddr into memory */ @@ -317,6 +482,15 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,if (ret <=0 ) goto put_old; +/* We are going to replace instruction, update ref_ctr. */ +if (!ref_ctr_updated && uprobe->ref_ctr_offset) { + ret =update_ref_ctr (uprobe, mm, is_register); +if (ret) + goto put_old; + + ref_ctr_updated =1 ; + } + ret =anon_vma_prepare (vma);if (ret) goto put_old;@@ -337,6 +511,11 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,if (unlikely (ret == -EAGAIN )) goto retry; + +/* Revert back reference counter if instruction update failed. */ +if (ret && is_register && ref_ctr_updated) +update_ref_ctr (uprobe, mm, false); +return ret; }@@ -378,8 +557,15 @@ static struct uprobe *get_uprobe(struct uprobe *uprobe)static void put_uprobe (struct uprobe *uprobe) { -if (atomic_dec_and_test (&uprobe->ref)) +if (atomic_dec_and_test (&uprobe->ref)) { +/* +* If application munmap(exec_vma) before uprobe_unregister() +* gets called, we don't get a chance to remove uprobe from +* delayed_uprobe_list in remove_breakpoint(). Do it here. +*/ +delayed_uprobe_remove (uprobe,NULL );kfree (uprobe); + } }static int match_uprobe (struct uprobe *l,struct uprobe *r)@@ -484,7 +670,8 @@ static struct uprobe *insert_uprobe(struct uprobe *uprobe)return u; } -static struct uprobe *alloc_uprobe (struct inode *inode,loff_t offset) +static struct uprobe *alloc_uprobe (struct inode *inode,loff_t offset, +loff_t ref_ctr_offset) {struct uprobe *uprobe, *cur_uprobe;@@ -494,6 +681,7 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset) uprobe->inode = inode; uprobe->offset = offset; + uprobe->ref_ctr_offset = ref_ctr_offset;init_rwsem (&uprobe->register_rwsem);init_rwsem (&uprobe->consumer_rwsem);@@ -895,7 +1083,7 @@ EXPORT_SYMBOL_GPL(uprobe_unregister);* else return 0 (success) */ static int __uprobe_register (struct inode *inode,loff_t offset, -struct uprobe_consumer *uc) +loff_t ref_ctr_offset,struct uprobe_consumer *uc) {struct uprobe *uprobe;int ret;@@ -912,7 +1100,7 @@ static int __uprobe_register(struct inode *inode, loff_t offset,return -EINVAL ; retry: - uprobe =alloc_uprobe (inode, offset); + uprobe =alloc_uprobe (inode, offset, ref_ctr_offset);if (!uprobe)return -ENOMEM ;/* @@ -938,10 +1126,17 @@ static int __uprobe_register(struct inode *inode, loff_t offset,int uprobe_register (struct inode *inode,loff_t offset,struct uprobe_consumer *uc) { -return __uprobe_register (inode, offset, uc); +return __uprobe_register (inode, offset,0 , uc); }EXPORT_SYMBOL_GPL (uprobe_register); +int uprobe_register_refctr (struct inode *inode,loff_t offset, +loff_t ref_ctr_offset,struct uprobe_consumer *uc) +{ +return __uprobe_register (inode, offset, ref_ctr_offset, uc); +} +EXPORT_SYMBOL_GPL (uprobe_register_refctr); +/* * uprobe_apply - unregister a already registered probe. * @inode: the file in which the probe has to be removed. kernel/events/uprobes.c static void build_probe_list(struct inode *inode, · line 1258
Quoted 4 lines
kernel/events/uprobes.c +1
-0@@ -1060,6 +1255,31 @@ static void build_probe_list(struct inode *inode,spin_unlock (&uprobes_treelock); } +static int delayed_uprobe_install (struct vm_area_struct *vma)This function name is confusing. How about we call it
delayed_ref_ctr_incr()or something similar? Also, we should add comments to highlight this is vma is not the vma containing the uprobe, but the vma containing theref_ctr.Quoted 8 lines
kernel/events/uprobes.c +8
-0+{ +struct list_head *pos, *q; +struct delayed_uprobe *du; +unsigned long vaddr; +int ret =0 , err =0 ; + +mutex_lock (&delayed_uprobe_lock); +list_for_each_safe (pos, q, &delayed_uprobe_list) {kernel/events/uprobes.c static void build_probe_list(struct inode *inode, · line 1276
Quoted 10 lines
kernel/events/uprobes.c +10
-0+ du = list_entry(pos, struct delayed_uprobe, list); + +if (!valid_ref_ctr_vma (du->uprobe, vma)) +continue ; + + vaddr =offset_to_vaddr (vma, du->uprobe->ref_ctr_offset); + ret =__update_ref_ctr (vma->vm_mm, vaddr,1 ); +/* Record an error and continue. */ +if (ret && !err) + err = ret;I think this is a good place (when ret != 0) to call
pr_warn(). I guess we can print which mm get error for which uprobe (inode+offset).Quoted 146 lines
kernel/events/uprobes.c +19
-1+delayed_uprobe_delete (du); + } +mutex_unlock (&delayed_uprobe_lock); +return err; +} +/* * Called from mmap_region/vma_adjust with mm->mmap_sem acquired. * @@ -1072,7 +1292,15 @@ int uprobe_mmap(struct vm_area_struct *vma)struct uprobe *uprobe, *u;struct inode *inode; -if (no_uprobe_events () || !valid_vma (vma, true)) +if (no_uprobe_events ()) +return 0 ; + +if (vma->vm_file && + (vma->vm_flags & (VM_WRITE |VM_SHARED )) ==VM_WRITE && +test_bit (MMF_HAS_UPROBES , &vma->vm_mm->flags)) +delayed_uprobe_install (vma); + +if (!valid_vma (vma, true))return 0 ; inode =file_inode (vma->vm_file);@@ -1246,6 +1474,10 @@ void uprobe_clear_state(struct mm_struct *mm) {struct xol_area *area = mm->uprobes_state.xol_area; +mutex_lock (&delayed_uprobe_lock); +delayed_uprobe_remove (NULL , mm); +mutex_unlock (&delayed_uprobe_lock); +if (!area)return ;kernel/trace/trace.c +1
-1@@ -4620,7 +4620,7 @@ static const char readme_msg[] = "place (kretprobe): [<module>:]<symbol>[+<offset>]|<memaddr>\n"#endif #ifdef CONFIG_UPROBE_EVENTS - "\t place: <path>:<offset>\n" + " place (uprobe): <path>:<offset>[(ref_ctr_offset)]\n"#endif "\t args: <name>=fetcharg[:type]\n" "\t fetcharg: %<register>, @<address>, @<symbol>[+|-<offset>],\n" kernel/trace/trace_uprobe.c +35
-4@@ -59,6 +59,7 @@ struct trace_uprobe {struct inode *inode;char *filename;unsigned long offset; +unsigned long ref_ctr_offset;unsigned long nhit;struct trace_probe tp; };@@ -364,10 +365,10 @@ static int register_trace_uprobe(struct trace_uprobe *tu)static int create_trace_uprobe (int argc,char **argv) {struct trace_uprobe *tu; -char *arg, *event, *group, *filename; +char *arg, *event, *group, *filename, *rctr, *rctr_end;char buf[MAX_EVENT_NAME_LEN ];struct path path; -unsigned long offset; +unsigned long offset, ref_ctr_offset;bool is_delete, is_return;int i, ret;@@ -376,6 +377,7 @@ static int create_trace_uprobe(int argc, char **argv) is_return = false; event =NULL ; group =NULL ; + ref_ctr_offset =0 ;/* argc must be >= 1 */ if (argv[0 ][0 ] =='-' )@@ -450,6 +452,26 @@ static int create_trace_uprobe(int argc, char **argv) goto fail_address_parse; } +/* Parse reference counter offset if specified. */ + rctr =strchr (arg,'(' ); +if (rctr) { + rctr_end =strchr (rctr,')' ); +if (rctr > rctr_end || *(rctr_end +1 ) !=0 ) { + ret = -EINVAL ; +pr_info ("Invalid reference counter offset.\n" ); + goto fail_address_parse; + } + + *rctr++ ='\0' ; + *rctr_end ='\0' ; + ret =kstrtoul (rctr,0 , &ref_ctr_offset); +if (ret) { +pr_info ("Invalid reference counter offset.\n" ); + goto fail_address_parse; + } + } + +/* Parse uprobe offset. */ ret =kstrtoul (arg,0 , &offset);if (ret) goto fail_address_parse;@@ -484,6 +506,7 @@ static int create_trace_uprobe(int argc, char **argv) goto fail_address_parse; } tu->offset = offset; + tu->ref_ctr_offset = ref_ctr_offset; tu->path = path; tu->filename =kstrdup (filename,GFP_KERNEL );@@ -602,6 +625,9 @@ static int probes_seq_show(struct seq_file *m, void *v)trace_event_name (&tu->tp.call), tu->filename, (int)(sizeof(void *) * 2), tu->offset); +if (tu->ref_ctr_offset) +seq_printf (m,"(0x%lx)" , tu->ref_ctr_offset); +for (i =0 ; i < tu->tp.nr_args; i++)seq_printf (m," %s=%s" , tu->tp.args[i].name, tu->tp.args[i].comm);@@ -917,7 +943,13 @@ probe_event_enable(struct trace_uprobe *tu, struct trace_event_file *file, tu->consumer.filter = filter; tu->inode =d_real_inode (tu->path.dentry); - ret =uprobe_register (tu->inode, tu->offset, &tu->consumer); +if (tu->ref_ctr_offset) { + ret =uprobe_register_refctr (tu->inode, tu->offset, + tu->ref_ctr_offset, &tu->consumer); + }else { + ret =uprobe_register (tu->inode, tu->offset, &tu->consumer); + } +if (ret) goto err_buffer; --2.14.4
Hi Song,
Quoted 6 lines · Song Liu
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 326
Quoted 5 lines
kernel/events/uprobes.c +5
-0+ +static void delayed_uprobe_delete (struct delayed_uprobe *du) +{ +if (!du) +return ;Do we really need this check?
Not necessary though, but I would still like to keep it for a safety.
Quoted 13 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 337
Quoted 12 lines
kernel/events/uprobes.c +11
-0+list_del (&du->list); +kfree (du); +} + +static void delayed_uprobe_remove (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct list_head *pos, *q; +struct delayed_uprobe *du; + +if (!uprobe && !mm) +return ;And do we really need this check?
Yes. delayed_uprobe_remove(uprobe=NULL, mm=NULL) is an invalid case. If I remove this check, code below (or more accurately code suggested by Oleg) will remove all entries from delayed_uprobe_list. So I will keep this check but put a comment above function.
[...]
Quoted 11 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 397
Quoted 11 lines
kernel/events/uprobes.c +11
-0+ + ret =get_user_pages_remote (NULL , mm, vaddr,1 , +FOLL_WRITE , &page, &vma,NULL ); +if (unlikely (ret <=0 )) { +/* +* We are asking for 1 page. If get_user_pages_remote() fails, +* it may return 0, in that case we have to return error. +*/ + ret = (ret ==0 ) ? -EBUSY : ret; +pr_warn ("Failed to %s ref_ctr. (%d)\n" , + d >0 ?"increment" :"decrement" , ret);
Quoted 3 lines
This warning is not really useful. Seems this function has little information about which uprobe is failing here. Maybe we only need warning in the caller (or caller of caller).
Sure, I can move this warning to caller of this function but what are the exact fields you would like to print with warning? Something like this is fine?
pr_warn("ref_ctr %s failed for 0x%lx, 0x%lx, 0x%lx, 0x%p",
d > 0 ? "increment" : "decrement", inode->i_ino,
offset, ref_ctr_offset, mm);More importantly, the reason I didn't print more info is because dmesg is accessible to unprivileged users in many distros but uprobes are not. So printing this information may be a security violation. No?
Quoted 18 lines
Quoted 18 lines
kernel/events/uprobes.c +17
-0+return ret; + } + + kaddr =kmap_atomic (page); + ptr = kaddr + (vaddr & ~PAGE_MASK ); + +if (unlikely (*ptr + d <0 )) { +pr_warn ("ref_ctr going negative. vaddr: 0x%lx, " +"curr val: %d, delta: %d\n" , vaddr, *ptr, d); + ret = -EINVAL ; + goto out; + } + + *ptr += d; + ret =0 ; +out: +kunmap_atomic (kaddr);
Quoted 8 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 420
Quoted 6 lines
kernel/events/uprobes.c +6
-0+put_page (page); +return ret; +} + +static int update_ref_ctr (struct uprobe *uprobe,struct mm_struct *mm, +bool is_register)What's the reason of bool
is_registerhere vs. short d in__update_ref_ctr()? Can we use short for both?
Yes, I can use short as well.
Quoted 10 lines
Quoted 10 lines
kernel/events/uprobes.c +9
-0+{ +struct vm_area_struct *rc_vma; +unsigned long rc_vaddr; +int ret =0 ; + + rc_vma =find_ref_ctr_vma (uprobe, mm); + +if (rc_vma) { + rc_vaddr =offset_to_vaddr (rc_vma, uprobe->ref_ctr_offset);
Quoted 8 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 434
Quoted 5 lines
kernel/events/uprobes.c +5
-0+ ret =__update_ref_ctr (mm, rc_vaddr, is_register ?1 :-1 ); + +if (is_register) +return ret; + }Mixing
__update_ref_ctr()here anddelayed_uprobe_add()in the same function is a little confusing (at least for me). How about we always use delayed uprobe foruprobe_mmap()and use non-delayed in other case(s)?
No. delayed_uprobe_add() is needed for uprobe_register() case to handle race between uprobe_register() and process creation.
[...]
Quoted 5 lines
Quoted 2 lines
+static int
delayed_uprobe_install(struct vm_area_struct *vma)This function name is confusing. How about we call it
delayed_ref_ctr_incr()or something similar? Also, we should add comments to highlight this is vma is not the vma containing the uprobe, but the vma containing theref_ctr.
Sure, I'll do that.
Quoted 13 lines
Quoted 13 lines
kernel/events/uprobes.c +12
-0+{ +struct list_head *pos, *q; +struct delayed_uprobe *du; +unsigned long vaddr; +int ret =0 , err =0 ; + +mutex_lock (&delayed_uprobe_lock); +list_for_each_safe (pos, q, &delayed_uprobe_list) { + du = list_entry(pos, struct delayed_uprobe, list); + +if (!valid_ref_ctr_vma (du->uprobe, vma)) +continue ;
Quoted 8 lines
kernel/events/uprobes.c static void build_probe_list(struct inode *inode, · line 1276
Quoted 6 lines
kernel/events/uprobes.c +6
-0+ + vaddr =offset_to_vaddr (vma, du->uprobe->ref_ctr_offset); + ret =__update_ref_ctr (vma->vm_mm, vaddr,1 ); +/* Record an error and continue. */ +if (ret && !err) + err = ret;I think this is a good place (when ret != 0) to call
pr_warn(). I guess we can print which mm get error for which uprobe (inode+offset).
__update_ref_ctr() is already printing warning, so I didn't add anything here. In case I remove a warning from __update_ref_ctr(), a warning something like below is fine?
pr_warn("ref_ctr increment failed for 0x%lx, 0x%lx, 0x%lx, 0x%p",
inode->i_ino, offset, ref_ctr_offset, vma->vm_mm);Again, can this lead to a security violation?
Thanks for detailed review :)
-Ravi
Hi Song,
Quoted 11 lines · Ravi Bangoria
Quoted 9 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 337
Quoted 8 lines
kernel/events/uprobes.c +8
-0+ +static void delayed_uprobe_remove (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct list_head *pos, *q; +struct delayed_uprobe *du; + +if (!uprobe && !mm) +return ;And do we really need this check?
Quoted 5 lines
Yes.
delayed_uprobe_remove(uprobe=NULL, mm=NULL)is an invalid case. If I remove this check, code below (or more accurately code suggested by Oleg) will remove all entries fromdelayed_uprobe_list. So I will keep this check but put a comment above function.
Sorry, my bad. Please ignore above comment. Even though, it saves us to unnecessary loop over entire delayed_uprobe_list when both uprobe and mm are NULL, that case is not possible with current code. Also, I'm not dereferencing any of them. So, IMHO it's fine to remove this check.
Thanks,
Ravi
Oleg Nesterov
maintainerQuoted 10 lines · Ravi Bangoria
Quoted 6 lines · Song Liu
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 326
Quoted 5 lines
kernel/events/uprobes.c +5
-0+ +static void delayed_uprobe_delete (struct delayed_uprobe *du) +{ +if (!du) +return ;Do we really need this check?
Not necessary though, but I would still like to keep it for a safety.
Heh. I tried to ignore all minor problems in this version, but now that Song mentioned this unnecessary check...
Personally I really dislike the checks like this one.
- It can confuse the reader who will try to understand the purpose
- it can hide a bug if delayed_uprobe_delete(du) is actually called with du == NULL.
IMO, you should either remove it and let the kernel crash (to notice the problem), or turn it into
if (WARN_ON(!du)) return;
which is self-documented and reports the problem without kernel crash.
Quoted 14 lines
Quoted 4 lines
Quoted 4 lines
kernel/events/uprobes.c +4
-0+ rc_vma =find_ref_ctr_vma (uprobe, mm); + +if (rc_vma) { + rc_vaddr =offset_to_vaddr (rc_vma, uprobe->ref_ctr_offset);Quoted 8 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 434
Quoted 5 lines
kernel/events/uprobes.c +5
-0+ ret =__update_ref_ctr (mm, rc_vaddr, is_register ?1 :-1 ); + +if (is_register) +return ret; + }Mixing
__update_ref_ctr()here anddelayed_uprobe_add()in the same function is a little confusing (at least for me). How about we always use delayed uprobe foruprobe_mmap()and use non-delayed in other case(s)?
Quoted 2 lines
No.
delayed_uprobe_add()is needed foruprobe_register()case to handle race betweenuprobe_register()and process creation.
Yes.
But damn, process creation (exec) is trivial. We could add a new uprobe_exec() hook and avoid delayed_uprobe_install() in uprobe_mmap().
Afaics, the really problematic case is dlopen() which can race with _register() too, right?
Oleg.
Hi Oleg,
Quoted 22 lines · Oleg Nesterov
Quoted 10 lines · Ravi Bangoria
Quoted 6 lines · Song Liu
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 326
Quoted 5 lines
kernel/events/uprobes.c +5
-0+ +static void delayed_uprobe_delete (struct delayed_uprobe *du) +{ +if (!du) +return ;Do we really need this check?
Not necessary though, but I would still like to keep it for a safety.
Heh. I tried to ignore all minor problems in this version, but now that Song mentioned this unnecessary check...
Personally I really dislike the checks like this one.
- It can confuse the reader who will try to understand the purpose
- it can hide a bug if
delayed_uprobe_delete(du)is actually called with du == NULL.
Quoted 7 lines
IMO, you should either remove it and let the kernel crash (to notice the problem), or turn it into
if (WARN_ON(!du)) return;which is self-documented and reports the problem without kernel crash.
Ok. I'll remove that check.
Quoted 15 lines
Quoted 15 lines
Quoted 13 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 434
Quoted 10 lines
kernel/events/uprobes.c +9
-0+ rc_vma =find_ref_ctr_vma (uprobe, mm); + +if (rc_vma) { + rc_vaddr =offset_to_vaddr (rc_vma, uprobe->ref_ctr_offset); + ret =__update_ref_ctr (mm, rc_vaddr, is_register ?1 :-1 ); + +if (is_register) +return ret; + }Mixing
__update_ref_ctr()here anddelayed_uprobe_add()in the same function is a little confusing (at least for me). How about we always use delayed uprobe foruprobe_mmap()and use non-delayed in other case(s)?
Quoted 7 lines
Quoted 2 lines
No.
delayed_uprobe_add()is needed foruprobe_register()case to handle race betweenuprobe_register()and process creation.Yes.
But damn, process creation (exec) is trivial. We could add a new
uprobe_exec()hook and avoiddelayed_uprobe_install()inuprobe_mmap().
I'm sorry. I didn't get this.
Quoted 3 lines
Afaics, the really problematic case is
dlopen()which can race with_register()too, right?
dlopen() should internally use mmap() right? So what is the problem here? Can you please elaborate.
Thanks,
Ravi
Oleg Nesterov
maintainerQuoted 5 lines · Ravi Bangoria
Quoted 3 lines
But damn, process creation (exec) is trivial. We could add a new
uprobe_exec()hook and avoiddelayed_uprobe_install()inuprobe_mmap().I'm sorry. I didn't get this.
Sorry for confusion...
I meant, if only exec*( could race with _register(), we could add another uprobe hook which updates all (delayed) counters before return to user-mode.
Quoted 5 lines
Quoted 2 lines
Afaics, the really problematic case is
dlopen()which can race with_register()too, right?
dlopen()should internally usemmap()right? So what is the problem here? Can you please elaborate.
What I tried to say is that we can't avoid uprobe_mmap()->delayed_uprobe_install() because dlopen() can race with _register() too, just like exec.
Oleg.
Quoted 11 lines · Oleg Nesterov
Quoted 5 lines · Ravi Bangoria
Quoted 3 lines
But damn, process creation (exec) is trivial. We could add a new
uprobe_exec()hook and avoiddelayed_uprobe_install()inuprobe_mmap().I'm sorry. I didn't get this.
Sorry for confusion...
I meant, if only exec*( could race with
_register(), we could add another uprobe hook which updates all (delayed) counters before return to user-mode.
Ok.
Quoted 9 lines
Quoted 6 lines
Quoted 3 lines
Afaics, the really problematic case is
dlopen()which can race with_register()too, right?
dlopen()should internally usemmap()right? So what is the problem here? Can you please elaborate.What I tried to say is that we can't avoid uprobe_mmap()->delayed_uprobe_install() because
dlopen()can race with_register()too, just like exec.
Right :)
Thanks,
Ravi
Oleg Nesterov
maintainerQuoted 13 lines · Oleg Nesterov
Quoted 5 lines · Ravi Bangoria
Quoted 3 lines
But damn, process creation (exec) is trivial. We could add a new
uprobe_exec()hook and avoiddelayed_uprobe_install()inuprobe_mmap().I'm sorry. I didn't get this.
Sorry for confusion...
I meant, if only exec*( could race with
_register(), we could add another uprobe hook which updates all (delayed) counters before return to user-mode.
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
Afaics, the really problematic case is
dlopen()which can race with_register()too, right?
dlopen()should internally usemmap()right? So what is the problem here? Can you please elaborate.What I tried to say is that we can't avoid uprobe_mmap()->delayed_uprobe_install() because
dlopen()can race with_register()too, just like exec.
I'm afraid I added even more confusion... so let me clarify although I am sure you understand this better than me ;)
of course, the main reason why we can't avoid uprobe_mmap()->delayed_uprobe_install() is not the race with _register(), the main reason is that dlopen() can mmap the code first, then mmap the counter we need to increment.
Again, just like exec, but exec is "atomic" in that it can't return to user-mode until evrything is done, so we could add another uprobe hook and avoid delayed_uprobe_install in uprobe_mmap().
Just in case my previous email was not clear.
Oleg.
Quoted 15 lines · Oleg Nesterov
Quoted 5 lines · Ravi Bangoria
Quoted 3 lines
But damn, process creation (exec) is trivial. We could add a new
uprobe_exec()hook and avoiddelayed_uprobe_install()inuprobe_mmap().I'm sorry. I didn't get this.
Sorry for confusion...
I meant, if only exec*( could race with
_register(), we could add another uprobe hook which updates all (delayed) counters before return to user-mode.Quoted 3 lines
Quoted 2 lines
Afaics, the really problematic case is
dlopen()which can race with_register()too, right?
Quoted 8 lines
Quoted 2 lines
dlopen()should internally usemmap()right? So what is the problem here? Can you please elaborate.What I tried to say is that we can't avoid uprobe_mmap()->delayed_uprobe_install() because
dlopen()can race with_register()too, just like exec.Oleg.
How about we do delayed_uprobe_install() per file? Say we keep a list of delayed_uprobe
in load_elf_binary(). Then we can install delayed_uprobe after loading all sections of the
file.
Song
Hi Song,
Quoted 22 lines · Song Liu
Quoted 21 lines · Oleg Nesterov
Quoted 5 lines · Ravi Bangoria
Quoted 3 lines
But damn, process creation (exec) is trivial. We could add a new
uprobe_exec()hook and avoiddelayed_uprobe_install()inuprobe_mmap().I'm sorry. I didn't get this.
Sorry for confusion...
I meant, if only exec*( could race with
_register(), we could add another uprobe hook which updates all (delayed) counters before return to user-mode.Quoted 5 lines
Quoted 2 lines
Afaics, the really problematic case is
dlopen()which can race with_register()too, right?
dlopen()should internally usemmap()right? So what is the problem here? Can you please elaborate.What I tried to say is that we can't avoid uprobe_mmap()->delayed_uprobe_install() because
dlopen()can race with_register()too, just like exec.
Quoted 8 lines
Quoted 2 lines
Oleg.
How about we do
delayed_uprobe_install()per file? Say we keep a list ofdelayed_uprobe
inload_elf_binary(). Then we can installdelayed_uprobeafter loading all sections of the
file.
I'm not sure if I totally understood the idea. But how this approach can solve dlopen() race with _register()?
Rather, making delayed_uprobe_list an mm field seems simple and effective idea to me. The only overhead will be list_empty(mm->delayed_list) check.
Please let me know if I misunderstood you.
Thanks,
Ravi
Quoted 37 lines · Ravi Bangoria
Hi Song,
Quoted 22 lines · Song Liu
Quoted 21 lines · Oleg Nesterov
Quoted 5 lines · Ravi Bangoria
Quoted 3 lines
But damn, process creation (exec) is trivial. We could add a new
uprobe_exec()hook and avoiddelayed_uprobe_install()inuprobe_mmap().I'm sorry. I didn't get this.
Sorry for confusion...
I meant, if only exec*( could race with
_register(), we could add another uprobe hook which updates all (delayed) counters before return to user-mode.Quoted 5 lines
Quoted 2 lines
Afaics, the really problematic case is
dlopen()which can race with_register()too, right?
dlopen()should internally usemmap()right? So what is the problem here? Can you please elaborate.What I tried to say is that we can't avoid uprobe_mmap()->delayed_uprobe_install() because
dlopen()can race with_register()too, just like exec.Quoted 8 lines
Quoted 2 lines
Oleg.
How about we do
delayed_uprobe_install()per file? Say we keep a list ofdelayed_uprobe
inload_elf_binary(). Then we can installdelayed_uprobeafter loading all sections of the
file.I'm not sure if I totally understood the idea. But how this approach can solve
dlopen()race with_register()?
Quoted 7 lines
Rather, making
delayed_uprobe_listan mm field seems simple and effective idea to me. The only overhead will belist_empty(mm->delayed_list)check.Please let me know if I misunderstood you.
Thanks,
Ravi
I misunderstood the problem here. I guess mm->delayed_list is the easiest solution of the race condition.
Thanks,
Song
Steven Rostedt
maintainerQuoted 19 lines · On Mon, 13 Aug 2018 13:50:19 +0200 Oleg Nesterov <[email protected]> wrote:
Quoted 10 lines · Ravi Bangoria
Quoted 6 lines · Song Liu
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 326
Quoted 5 lines
kernel/events/uprobes.c +5
-0+ +static void delayed_uprobe_delete (struct delayed_uprobe *du) +{ +if (!du) +return ;Do we really need this check?
Not necessary though, but I would still like to keep it for a safety.
Heh. I tried to ignore all minor problems in this version, but now that Song mentioned this unnecessary check...
Personally I really dislike the checks like this one.
- It can confuse the reader who will try to understand the purpose
Quoted 8 lines
- it can hide a bug if
delayed_uprobe_delete(du)is actually called with du == NULL.IMO, you should either remove it and let the kernel crash (to notice the problem), or turn it into
if (WARN_ON(!du)) return;
I'd prefer the more robust WARN_ON(!du) above instead of removing it.
-- Steve
Quoted 116 lines · Ravi Bangoria
Hi Song,
Quoted 6 lines · Song Liu
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 326
Quoted 5 lines
kernel/events/uprobes.c +5
-0+ +static void delayed_uprobe_delete (struct delayed_uprobe *du) +{ +if (!du) +return ;Do we really need this check?
Not necessary though, but I would still like to keep it for a safety.
Quoted 13 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 337
Quoted 12 lines
kernel/events/uprobes.c +11
-0+list_del (&du->list); +kfree (du); +} + +static void delayed_uprobe_remove (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct list_head *pos, *q; +struct delayed_uprobe *du; + +if (!uprobe && !mm) +return ;And do we really need this check?
Yes.
delayed_uprobe_remove(uprobe=NULL, mm=NULL)is an invalid case. If I remove this check, code below (or more accurately code suggested by Oleg) will remove all entries fromdelayed_uprobe_list. So I will keep this check but put a comment above function.[...]
Quoted 11 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 397
Quoted 11 lines
kernel/events/uprobes.c +11
-0+ + ret =get_user_pages_remote (NULL , mm, vaddr,1 , +FOLL_WRITE , &page, &vma,NULL ); +if (unlikely (ret <=0 )) { +/* +* We are asking for 1 page. If get_user_pages_remote() fails, +* it may return 0, in that case we have to return error. +*/ + ret = (ret ==0 ) ? -EBUSY : ret; +pr_warn ("Failed to %s ref_ctr. (%d)\n" , + d >0 ?"increment" :"decrement" , ret);Quoted 3 lines
This warning is not really useful. Seems this function has little information about which uprobe is failing here. Maybe we only need warning in the caller (or caller of caller).
Sure, I can move this warning to caller of this function but what are the exact fields you would like to print with warning? Something like this is fine?
pr_warn("ref_ctr %s failed for 0x%lx, 0x%lx, 0x%lx, 0x%p", d > 0 ? "increment" : "decrement", inode->i_ino, offset, ref_ctr_offset, mm);More importantly, the reason I didn't print more info is because dmesg is accessible to unprivileged users in many distros but uprobes are not. So printing this information may be a security violation. No?
Quoted 18 lines
Quoted 18 lines
kernel/events/uprobes.c +17
-0+return ret; + } + + kaddr =kmap_atomic (page); + ptr = kaddr + (vaddr & ~PAGE_MASK ); + +if (unlikely (*ptr + d <0 )) { +pr_warn ("ref_ctr going negative. vaddr: 0x%lx, " +"curr val: %d, delta: %d\n" , vaddr, *ptr, d); + ret = -EINVAL ; + goto out; + } + + *ptr += d; + ret =0 ; +out: +kunmap_atomic (kaddr);Quoted 8 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 420
Quoted 6 lines
kernel/events/uprobes.c +6
-0+put_page (page); +return ret; +} + +static int update_ref_ctr (struct uprobe *uprobe,struct mm_struct *mm, +bool is_register)What's the reason of bool
is_registerhere vs. short d in__update_ref_ctr()? Can we use short for both?Yes, I can use short as well.
Quoted 10 lines
Quoted 10 lines
kernel/events/uprobes.c +9
-0+{ +struct vm_area_struct *rc_vma; +unsigned long rc_vaddr; +int ret =0 ; + + rc_vma =find_ref_ctr_vma (uprobe, mm); + +if (rc_vma) { + rc_vaddr =offset_to_vaddr (rc_vma, uprobe->ref_ctr_offset);Quoted 8 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 434
Quoted 5 lines
kernel/events/uprobes.c +5
-0+ ret =__update_ref_ctr (mm, rc_vaddr, is_register ?1 :-1 ); + +if (is_register) +return ret; + }Mixing
__update_ref_ctr()here anddelayed_uprobe_add()in the same function is a little confusing (at least for me). How about we always use delayed uprobe foruprobe_mmap()and use non-delayed in other case(s)?
Quoted 2 lines
No.
delayed_uprobe_add()is needed foruprobe_register()case to handle race betweenuprobe_register()and process creation.
I see.
Quoted 36 lines
[...]
Quoted 5 lines
Quoted 2 lines
+static int
delayed_uprobe_install(struct vm_area_struct *vma)This function name is confusing. How about we call it
delayed_ref_ctr_incr()or something similar? Also, we should add comments to highlight this is vma is not the vma containing the uprobe, but the vma containing theref_ctr.Sure, I'll do that.
Quoted 13 lines
Quoted 13 lines
kernel/events/uprobes.c +12
-0+{ +struct list_head *pos, *q; +struct delayed_uprobe *du; +unsigned long vaddr; +int ret =0 , err =0 ; + +mutex_lock (&delayed_uprobe_lock); +list_for_each_safe (pos, q, &delayed_uprobe_list) { + du = list_entry(pos, struct delayed_uprobe, list); + +if (!valid_ref_ctr_vma (du->uprobe, vma)) +continue ;Quoted 8 lines
kernel/events/uprobes.c static void build_probe_list(struct inode *inode, · line 1276
Quoted 6 lines
kernel/events/uprobes.c +6
-0+ + vaddr =offset_to_vaddr (vma, du->uprobe->ref_ctr_offset); + ret =__update_ref_ctr (vma->vm_mm, vaddr,1 ); +/* Record an error and continue. */ +if (ret && !err) + err = ret;I think this is a good place (when ret != 0) to call
pr_warn(). I guess we can print which mm get error for which uprobe (inode+offset).
Quoted 7 lines
__update_ref_ctr()is already printing warning, so I didn't add anything here. In case I remove a warning from__update_ref_ctr(), a warning something like below is fine?pr_warn("ref_ctr increment failed for 0x%lx, 0x%lx, 0x%lx, 0x%p", inode->i_ino, offset, ref_ctr_offset, vma->vm_mm);
I was thinking about a message like:
ref_ctr increment failed for inode XX offset YY ref_ctr ZZ of mm 0xWWW
I didn't thought about the security part of it, but I guess it is OK.
Thanks,
Song
Quoted 5 lines
Again, can this lead to a security violation?
Thanks for detailed review :)
-Ravi
Quoted 8 lines
kernel/events/uprobes.c +8
-0+ +static int delayed_uprobe_add (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct delayed_uprobe *du; + +if (delayed_uprobe_check (uprobe, mm)) +return 0 ; +
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 322
Quoted 10 lines
kernel/events/uprobes.c +10
-0+ du =kzalloc (sizeof (*du),GFP_KERNEL ); +if (!du) +return -ENOMEM ; + + du->uprobe = uprobe; + du->mm = mm; +list_add (&du->list, &delayed_uprobe_list); +return 0 ; +} +
Should we keep the delayed uprobe list per mm?
That way we could avoid the global mutex lock.
And the list to traverse would also be small.
kernel/events/uprobes.c static struct uprobe *get_uprobe(struct uprobe *uprobe) · line 566
Quoted 13 lines
kernel/events/uprobes.c +7
-1@@ -378,8 +557,15 @@ static struct uprobe *get_uprobe(struct uprobe *uprobe)static void put_uprobe (struct uprobe *uprobe) { -if (atomic_dec_and_test (&uprobe->ref)) +if (atomic_dec_and_test (&uprobe->ref)) { +/* +* If application munmap(exec_vma) before uprobe_unregister() +* gets called, we don't get a chance to remove uprobe from +* delayed_uprobe_list in remove_breakpoint(). Do it here. +*/ +delayed_uprobe_remove (uprobe,NULL );
I am not getting this part. If unmap happens before unregister, why cant we use uprobe_munmap? what am I missing.
Quoted 4 lines
kernel/events/uprobes.c +1
-0kfree (uprobe); + } }
Oleg Nesterov
maintainerQuoted 15 lines · Srikar Dronamraju
Quoted 15 lines
kernel/events/uprobes.c +14
-0+ +static int delayed_uprobe_add (struct uprobe *uprobe,struct mm_struct *mm) +{ +struct delayed_uprobe *du; + +if (delayed_uprobe_check (uprobe, mm)) +return 0 ; + + du =kzalloc (sizeof (*du),GFP_KERNEL ); +if (!du) +return -ENOMEM ; + + du->uprobe = uprobe; + du->mm = mm;
Quoted 8 lines
kernel/events/uprobes.c static int verify_opcode(struct page *page, unsigned long vaddr, uprobe_opcode_t · line 322
Quoted 4 lines
kernel/events/uprobes.c +4
-0+list_add (&du->list, &delayed_uprobe_list); +return 0 ; +} +Should we keep the delayed uprobe list per mm?
That way we could avoid the global mutex lock.
And the list to traverse would also be small.
Plus uprobe_mmap() could simply check list_empty(mm->delayed_list) before the costly delayed_uprobe_install().
Yes, I mentioned this too. But then I suggested to not do this in the initial version to make it more simple.
Hopefully we can do this later, but note that this conflicts with the change in put_uprobe() you commented below.
Quoted 5 lines
Quoted 5 lines
static void
put_uprobe(struct uprobe *uprobe)
{kernel/events/uprobes.c +2
-1-if (atomic_dec_and_test (&uprobe->ref)) +if (atomic_dec_and_test (&uprobe->ref)) { +/*
Quoted 8 lines
kernel/events/uprobes.c static struct uprobe *get_uprobe(struct uprobe *uprobe) · line 566
Quoted 5 lines
kernel/events/uprobes.c +5
-0+* If application munmap(exec_vma) before uprobe_unregister() +* gets called, we don't get a chance to remove uprobe from +* delayed_uprobe_list in remove_breakpoint(). Do it here. +*/ +delayed_uprobe_remove (uprobe,NULL );I am not getting this part. If unmap happens before unregister, why cant we use
uprobe_munmap?
How? I mean what exactly can we do in uprobe_munmap() ? Firstly, we will need to restore build_probe_list/list_for_each_entry_safe(pending_list) in uprobe_munmap() and this is not nice performance-wise. Then what?
We don't even know if the caller is actually munmap(), it can be vma_adjust() and in the latter case we can't do delayed_uprobe_remove(uprobe, mm).
Perhaps we can use uprobe_munmap() to cleanup the delayed_uprobe_list, but this doesn't look simple to me.
In fact I think that we should simply kill uprobe_munmap().
Oleg.
Quoted 4 lines
kernel/events/uprobes.c +3
-0+static int delayed_uprobe_install (struct vm_area_struct *vma) +{ +struct list_head *pos, *q;
kernel/events/uprobes.c static void build_probe_list(struct inode *inode, · line 1270
Quoted 10 lines
kernel/events/uprobes.c +10
-0+struct delayed_uprobe *du; +unsigned long vaddr; +int ret =0 , err =0 ; + +mutex_lock (&delayed_uprobe_lock); +list_for_each_safe (pos, q, &delayed_uprobe_list) { + du = list_entry(pos, struct delayed_uprobe, list); + +if (!valid_ref_ctr_vma (du->uprobe, vma)) +continue ;
I think we should compare mm here. I.e.:
if (du->mm != vma->vm_mm || !valid_ref_ctr_vma(du->uprobe, vma))
continue;Otherwise things can mess up.
Quoted 11 lines
kernel/events/uprobes.c +11
-0+ + vaddr =offset_to_vaddr (vma, du->uprobe->ref_ctr_offset); + ret =__update_ref_ctr (vma->vm_mm, vaddr,1 ); +/* Record an error and continue. */ +if (ret && !err) + err = ret; +delayed_uprobe_delete (du); + } +mutex_unlock (&delayed_uprobe_lock); +return err; +}
PATCH v8 4/6 Uprobes/sdt: Prevent multiple reference counter for same uprobe
In reply to Ravi BangoriaWe assume to have only one reference counter for one uprobe. Don't allow user to register multiple uprobes having same inode+offset but different reference counter.
--- kernel/events/uprobes.c | 9 +++++++++ 1 file changed, 9 insertions(+)
kernel/events/uprobes.c +9 -0
@@ -689,6 +689,12 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset, cur_uprobe =insert_uprobe (uprobe);/* a uprobe exists for this inode:offset combination */ if (cur_uprobe) { +if (cur_uprobe->ref_ctr_offset != uprobe->ref_ctr_offset) { +pr_warn ("Reference counter offset mismatch.\n" ); +put_uprobe (cur_uprobe); +kfree (uprobe); +return ERR_PTR (-EINVAL ); + }kfree (uprobe); uprobe = cur_uprobe; }
@@ -1103,6 +1109,9 @@ static int __uprobe_register(struct inode *inode, loff_t offset, uprobe =alloc_uprobe (inode, offset, ref_ctr_offset);if (!uprobe)return -ENOMEM ; +if (IS_ERR (uprobe)) +return PTR_ERR (uprobe); +/* * We can race with uprobe_unregister()->delete_uprobe(). * Check uprobe_is_active() and retry if it is false.
Signature · 1 line
2.14.4
Re: [PATCH v8 4/6] Uprobes/sdt: Prevent multiple reference counter for same uprobe
Quoted 14 lines · Ravi Bangoria
We assume to have only one reference counter for one uprobe. Don't allow user to register multiple uprobes having same inode+offset but different reference counter.
Signed-off-by Ravi Bangoria <[email protected]>--- kernel/events/uprobes.c | 9 +++++++++ 1 file changed, 9 insertions(+)kernel/events/uprobes.c +0
-0
kernel/events/uprobes.c static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset, · line 693
Quoted 5 lines
kernel/events/uprobes.c +2
-0@@ -689,6 +689,12 @@ static struct uprobe *alloc_uprobe(struct inode *inode, loff_t offset, cur_uprobe =insert_uprobe (uprobe);/* a uprobe exists for this inode:offset combination */ if (cur_uprobe) { +if (cur_uprobe->ref_ctr_offset != uprobe->ref_ctr_offset) { +pr_warn ("Reference counter offset mismatch.\n" );
We should print more information here, including the inode, the offset and both ref_ctr_offset.
Quoted 20 lines
kernel/events/uprobes.c +7
-1+put_uprobe (cur_uprobe); +kfree (uprobe); +return ERR_PTR (-EINVAL ); + }kfree (uprobe); uprobe = cur_uprobe; }@@ -1103,6 +1109,9 @@ static int __uprobe_register(struct inode *inode, loff_t offset, uprobe =alloc_uprobe (inode, offset, ref_ctr_offset);if (!uprobe)return -ENOMEM ; +if (IS_ERR (uprobe)) +return PTR_ERR (uprobe); +/* * We can race with uprobe_unregister()->delete_uprobe(). * Check uprobe_is_active() and retry if it is false. -- 2.14.4
Re: [PATCH v8 4/6] Uprobes/sdt: Prevent multiple reference counter for same uprobe
In reply to Ravi BangoriaQuoted 6 lines · Ravi Bangoria
We assume to have only one reference counter for one uprobe. Don't allow user to register multiple uprobes having same inode+offset but different reference counter.
Signed-off-by Ravi Bangoria <[email protected]>---
Looks good to me.
PATCH v8 5/6 trace_uprobe/sdt: Prevent multiple reference counter for same uprobe
In reply to Ravi BangoriaWe assume to have only one reference counter for one uprobe. Don't allow user to add multiple trace_uprobe entries having same inode+offset but different reference counter.
Ex,
# echo "p:sdt_tick/loop2 /home/ravi/tick:0x6e4(0x10036)" > uprobe_events # echo "p:sdt_tick/loop2_1 /home/ravi/tick:0x6e4(0xfffff)" >> uprobe_events bash: echo: write error: Invalid argument
# dmesg trace_kprobe: Reference counter offset mismatch.
There is one exception though:
When user is trying to replace the old entry with the new one, we allow this if the new entry does not conflict with any other existing entries.
--- kernel/trace/trace_uprobe.c | 37 +++++++++++++++++++++++++++++++++++-- 1 file changed, 35 insertions(+), 2 deletions(-)
kernel/trace/trace_uprobe.c +35 -2
@@ -324,6 +324,35 @@ static int unregister_trace_uprobe(struct trace_uprobe *tu)return 0 ; } +/* +* Uprobe with multiple reference counter is not allowed. i.e. +* If inode and offset matches, reference counter offset *must* +* match as well. Though, there is one exception: If user is +* replacing old trace_uprobe with new one(same group/event), +* then we allow same uprobe with new reference counter as far +* as the new one does not conflict with any other existing +* ones. +*/ +static struct trace_uprobe *find_old_trace_uprobe (struct trace_uprobe *new) +{ +struct trace_uprobe *tmp, *old =NULL ; +struct inode *new_inode =d_real_inode (new->path.dentry); + + old =find_probe_event (trace_event_name (&new->tp.call), + new->tp.call.class->system); + +list_for_each_entry (tmp, &uprobe_list, list) { +if ((old ? old != tmp : true) && + new_inode ==d_real_inode (tmp->path.dentry) && + new->offset == tmp->offset && + new->ref_ctr_offset != tmp->ref_ctr_offset) { +pr_warn ("Reference counter offset mismatch." ); +return ERR_PTR (-EINVAL ); + } + } +return old; +} +/* Register a trace_uprobe and probe_event */ static int register_trace_uprobe (struct trace_uprobe *tu) {
@@ -333,8 +362,12 @@ static int register_trace_uprobe(struct trace_uprobe *tu)mutex_lock (&uprobe_lock);/* register as an event */ - old_tu =find_probe_event (trace_event_name (&tu->tp.call), - tu->tp.call.class->system); + old_tu =find_old_trace_uprobe (tu); +if (IS_ERR (old_tu)) { + ret =PTR_ERR (old_tu); + goto end; + } +if (old_tu) {/* delete old event */ ret =unregister_trace_uprobe (old_tu);
Signature · 1 line
2.14.4
Re: [PATCH v8 5/6] trace_uprobe/sdt: Prevent multiple reference counter for same uprobe
Do we really need this given we already have PATCH 4/6? uprobe_regsiter() can be called
out of trace_uprobe, this patch won't catch all conflicts anyway.
Song
Quoted 80 lines · Ravi Bangoria
We assume to have only one reference counter for one uprobe. Don't allow user to add multiple
trace_uprobeentries having same inode+offset but different reference counter.Ex,
# echo "p:sdt_tick/loop2 /home/ravi/tick:0x6e4(0x10036)" > uprobe_events # echo "p:sdt_tick/loop2_1 /home/ravi/tick:0x6e4(0xfffff)" >> uprobe_events bash: echo: write error: Invalid argument# dmesg trace_kprobe: Reference counter offset mismatch.There is one exception though:
When user is trying to replace the old entry with the new one, we allow this if the new entry does not conflict with any other existing entries.Signed-off-by Ravi Bangoria <[email protected]>--- kernel/trace/trace_uprobe.c | 37 +++++++++++++++++++++++++++++++++++-- 1 file changed, 35 insertions(+), 2 deletions(-)kernel/trace/trace_uprobe.c +35
-3@@ -324,6 +324,35 @@ static int unregister_trace_uprobe(struct trace_uprobe *tu)return 0 ; } +/* +* Uprobe with multiple reference counter is not allowed. i.e. +* If inode and offset matches, reference counter offset *must* +* match as well. Though, there is one exception: If user is +* replacing old trace_uprobe with new one(same group/event), +* then we allow same uprobe with new reference counter as far +* as the new one does not conflict with any other existing +* ones. +*/ +static struct trace_uprobe *find_old_trace_uprobe (struct trace_uprobe *new) +{ +struct trace_uprobe *tmp, *old =NULL ; +struct inode *new_inode =d_real_inode (new->path.dentry); + + old =find_probe_event (trace_event_name (&new->tp.call), + new->tp.call.class->system); + +list_for_each_entry (tmp, &uprobe_list, list) { +if ((old ? old != tmp : true) && + new_inode ==d_real_inode (tmp->path.dentry) && + new->offset == tmp->offset && + new->ref_ctr_offset != tmp->ref_ctr_offset) { +pr_warn ("Reference counter offset mismatch." ); +return ERR_PTR (-EINVAL ); + } + } +return old; +} +/* Register a trace_uprobe and probe_event */ static int register_trace_uprobe (struct trace_uprobe *tu) {@@ -333,8 +362,12 @@ static int register_trace_uprobe(struct trace_uprobe *tu)mutex_lock (&uprobe_lock);/* register as an event */ - old_tu =find_probe_event (trace_event_name (&tu->tp.call), - tu->tp.call.class->system); + old_tu =find_old_trace_uprobe (tu); +if (IS_ERR (old_tu)) { + ret =PTR_ERR (old_tu); + goto end; + } +if (old_tu) {/* delete old event */ ret =unregister_trace_uprobe (old_tu); --2.14.4
Re: [PATCH v8 5/6] trace_uprobe/sdt: Prevent multiple reference counter for same uprobe
Hi Song,
Quoted 3 lines · Song Liu
Do we really need this given we already have PATCH 4/6?
uprobe_regsiter()can be called
out oftrace_uprobe, this patch won't catch all conflicts anyway.
Right but it, at least, catch all conflicts happening via trace_uprobe.
I don't mind in removing this patch but I would like to get an opinion of Oleg/Srikar/Steven/Masami.
Thanks,
Ravi
Re: [PATCH v8 5/6] trace_uprobe/sdt: Prevent multiple reference counter for same uprobe
Quoted 12 lines · Ravi Bangoria
Hi Song,
Quoted 3 lines · Song Liu
Do we really need this given we already have PATCH 4/6?
uprobe_regsiter()can be called
out oftrace_uprobe, this patch won't catch all conflicts anyway.Right but it, at least, catch all conflicts happening via
trace_uprobe.I don't mind in removing this patch but I would like to get an opinion of Oleg/Srikar/Steven/Masami.
I would suggest to keep it, atleast it can ctah conflicts happening via trace_uprobe.
Song Liu
Reviewed-byRe: [PATCH v8 5/6] trace_uprobe/sdt: Prevent multiple reference counter for same uprobe
Quoted 9 lines · Srikar Dronamraju
Quoted 7 lines · Ravi Bangoria
Hi Song,
Quoted 3 lines · Song Liu
Do we really need this given we already have PATCH 4/6?
uprobe_regsiter()can be called
out oftrace_uprobe, this patch won't catch all conflicts anyway.
Quoted 9 lines
Quoted 5 lines
Right but it, at least, catch all conflicts happening via
trace_uprobe.I don't mind in removing this patch but I would like to get an opinion of Oleg/Srikar/Steven/Masami.
I would suggest to keep it, atleast it can ctah conflicts happening via
trace_uprobe.
Yeah, that makes sense.
Re: [PATCH v8 5/6] trace_uprobe/sdt: Prevent multiple reference counter for same uprobe
In reply to Ravi BangoriaQuoted 12 lines · Ravi Bangoria
We assume to have only one reference counter for one uprobe. Don't allow user to add multiple
trace_uprobeentries having same inode+offset but different reference counter.Ex,
# echo "p:sdt_tick/loop2 /home/ravi/tick:0x6e4(0x10036)" > uprobe_events # echo "p:sdt_tick/loop2_1 /home/ravi/tick:0x6e4(0xfffff)" >> uprobe_events bash: echo: write error: Invalid argument# dmesg trace_kprobe: Reference counter offset mismatch.
Quoted 7 lines
There is one exception though:
When user is trying to replace the old entry with the new one, we allow this if the new entry does not conflict with any other existing entries.Signed-off-by Ravi Bangoria <[email protected]>---
Looks good to me.
PATCH v8 6/6 perf probe: Support SDT markers having reference counter (semaphore)
In reply to Ravi BangoriaWith this, perf buildid-cache will save SDT markers with reference counter in probe cache. Perf probe will be able to probe markers having reference counter. Ex,
# readelf -n /tmp/tick | grep -A1 loop2
Name: loop2
... Semaphore: 0x0000000010020036 # ./perf buildid-cache --add /tmp/tick
# ./perf probe sdt_tick:loop2
# ./perf stat -e sdt_tick:loop2 /tmp/tick
hi: 0
hi: 1
hi: 2
^C
Performance counter stats for '/tmp/tick':
3 sdt_tick:loop2
2.561851452 seconds time elapsed--- tools/perf/util/probe-event.c | 39 ++++++++++++++++++++++++++++++++---- tools/perf/util/probe-event.h | 1 + tools/perf/util/probe-file.c | 34 ++++++++++++++++++++++++++------ tools/perf/util/probe-file.h | 1 + tools/perf/util/symbol-elf.c | 46 ++++++++++++++++++++++++++++++++----------- tools/perf/util/symbol.h | 7 +++++++ 6 files changed, 106 insertions(+), 22 deletions(-)
tools/perf/util/probe-event.c +35 -4
@@ -1819,6 +1819,12 @@ int parse_probe_trace_command(const char *cmd, struct probe_trace_event *tev) tp->offset =strtoul (fmt2_str,NULL ,10 ); } +if (tev->uprobes) { + fmt2_str =strchr (p,'(' ); +if (fmt2_str) + tp->ref_ctr_offset =strtoul (fmt2_str +1 ,NULL ,0 ); + } + tev->nargs = argc -2 ; tev->args =zalloc (sizeof (struct probe_trace_arg ) * tev->nargs);if (tev->args ==NULL ) {
@@ -2012,6 +2018,22 @@ static int synthesize_probe_trace_arg(struct probe_trace_arg *arg,return err; } +static int +synthesize_uprobe_trace_def (struct probe_trace_event *tev,struct strbuf *buf) +{ +struct probe_trace_point *tp = &tev->point; +int err; + + err =strbuf_addf (buf,"%s:0x%lx" , tp->module, tp->address); + +if (err >=0 && tp->ref_ctr_offset) { +if (!uprobe_ref_ctr_is_supported ()) +return -1 ; + err =strbuf_addf (buf,"(0x%lx)" , tp->ref_ctr_offset); + } +return err >=0 ?0 :-1 ; +} +char *synthesize_probe_trace_command (struct probe_trace_event *tev) {struct probe_trace_point *tp = &tev->point;
@@ -2041,15 +2063,17 @@ char *synthesize_probe_trace_command(struct probe_trace_event *tev) }/* Use the tp->address for uprobes */ -if (tev->uprobes) - err =strbuf_addf (&buf,"%s:0x%lx" , tp->module, tp->address); -else if (!strncmp (tp->symbol,"0x" ,2 )) +if (tev->uprobes) { + err =synthesize_uprobe_trace_def (tev, &buf); + }else if (!strncmp (tp->symbol,"0x" ,2 )) {/* Absolute address. See try_to_find_absolute_address() */ err =strbuf_addf (&buf,"%s%s0x%lx" , tp->module ?:"" , tp->module ?":" :"" , tp->address); -else + }else { err =strbuf_addf (&buf,"%s%s%s+%lu" , tp->module ?:"" , tp->module ?":" :"" , tp->symbol, tp->offset); + } +if (err) goto error;
@@ -2633,6 +2657,13 @@ static void warn_uprobe_event_compat(struct probe_trace_event *tev) {int i;char *buf =synthesize_probe_trace_command (tev); +struct probe_trace_point *tp = &tev->point; + +if (tp->ref_ctr_offset && !uprobe_ref_ctr_is_supported ()) { +pr_warning ("A semaphore is associated with %s:%s and " +"seems your kernel doesn't support it.\n" , + tev->group, tev->event); + }/* Old uprobe event doesn't support memory dereference */ if (!tev->uprobes || tev->nargs ==0 || !buf)
tools/perf/util/probe-event.h +1 -0
@@ -27,6 +27,7 @@ struct probe_trace_point {char *symbol;/* Base symbol */ char *module;/* Module name */ unsigned long offset;/* Offset from symbol */ +unsigned long ref_ctr_offset;/* SDT reference counter offset */ unsigned long address;/* Actual address of the trace point */ bool retprobe;/* Return probe flag */ };
tools/perf/util/probe-file.c +28 -6
@@ -696,8 +696,16 @@ int probe_cache__add_entry(struct probe_cache *pcache,#ifdef HAVE_GELF_GETNOTE_SUPPORT static unsigned long long sdt_note__get_addr (struct sdt_note *note) { -return note->bit32 ? (unsigned long long )note->addr.a32[0 ] - : (unsigned long long )note->addr.a64[0 ]; +return note->bit32 ? + (unsigned long long )note->addr.a32[SDT_NOTE_IDX_LOC ] : + (unsigned long long )note->addr.a64[SDT_NOTE_IDX_LOC ]; +} + +static unsigned long long sdt_note__get_ref_ctr_offset (struct sdt_note *note) +{ +return note->bit32 ? + (unsigned long long )note->addr.a32[SDT_NOTE_IDX_REFCTR ] : + (unsigned long long )note->addr.a64[SDT_NOTE_IDX_REFCTR ]; } static const char * const type_to_suffix[] = {
@@ -775,14 +783,21 @@ static char *synthesize_sdt_probe_command(struct sdt_note *note, {struct strbuf buf;char *ret =NULL , **args; -int i, args_count; +int i, args_count, err; +unsigned long long ref_ctr_offset;if (strbuf_init (&buf,32 ) <0 )return NULL ; -if (strbuf_addf (&buf,"p:%s/%s %s:0x%llx" , - sdtgrp, note->name, pathname, -sdt_note__get_addr (note)) <0 ) + err =strbuf_addf (&buf,"p:%s/%s %s:0x%llx" , + sdtgrp, note->name, pathname, +sdt_note__get_addr (note)); + + ref_ctr_offset =sdt_note__get_ref_ctr_offset (note); +if (ref_ctr_offset && err >=0 ) + err =strbuf_addf (&buf,"(0x%llx)" , ref_ctr_offset); + +if (err <0 ) goto error;if (!note->args)
@@ -998,6 +1013,7 @@ int probe_cache__show_all_caches(struct strfilter *filter)enum ftrace_readme {FTRACE_README_PROBE_TYPE_X =0 ,FTRACE_README_KRETPROBE_OFFSET , +FTRACE_README_UPROBE_REF_CTR ,FTRACE_README_END , };
@@ -1009,6 +1025,7 @@ static struct { [idx] = {.pattern = pat, .avail = false}DEFINE_TYPE (FTRACE_README_PROBE_TYPE_X ,"*type: * x8/16/32/64,*" ),DEFINE_TYPE (FTRACE_README_KRETPROBE_OFFSET ,"*place (kretprobe): *" ), + DEFINE_TYPE(FTRACE_README_UPROBE_REF_CTR, "*ref_ctr_offset*"), };static bool scan_ftrace_readme (enum ftrace_readme type)
@@ -1064,3 +1081,8 @@ bool kretprobe_offset_is_supported(void) {return scan_ftrace_readme (FTRACE_README_KRETPROBE_OFFSET ); } + +bool uprobe_ref_ctr_is_supported (void ) +{ +return scan_ftrace_readme (FTRACE_README_UPROBE_REF_CTR ); +}
tools/perf/util/probe-file.h +1 -0
@@ -69,6 +69,7 @@ struct probe_cache_entry *probe_cache__find_by_name(struct probe_cache *pcache,int probe_cache__show_all_caches (struct strfilter *filter);bool probe_type_is_available (enum probe_type type);bool kretprobe_offset_is_supported (void ); +bool uprobe_ref_ctr_is_supported (void );#else /* ! HAVE_LIBELF_SUPPORT */ static inline struct probe_cache *probe_cache__new (const char *tgt__maybe_unused ,struct nsinfo *nsi__maybe_unused ) {
tools/perf/util/symbol-elf.c +34 -12
64 lines not on this page · raw message
tools/perf/util/symbol.h +7 -0
@@ -379,12 +379,19 @@ int get_sdt_note_list(struct list_head *head, const char *target);int cleanup_sdt_note_list (struct list_head *sdt_notes);int sdt_notes__get_count (struct list_head *start); +#define SDT_PROBES_SCN ".probes"#define SDT_BASE_SCN ".stapsdt.base"#define SDT_NOTE_SCN ".note.stapsdt"#define SDT_NOTE_TYPE 3#define SDT_NOTE_NAME "stapsdt"#define NR_ADDR 3 +enum { +SDT_NOTE_IDX_LOC =0 , +SDT_NOTE_IDX_BASE , +SDT_NOTE_IDX_REFCTR , +}; +struct mem_info *mem_info__new (void );struct mem_info *mem_info__get (struct mem_info *mi);void mem_info__put (struct mem_info *mi);
Signature · 1 line
2.14.4
Quoted 7 lines · Ravi Bangoria
Ravi Bangoria (6):
Uprobes: Simplifyuprobe_register()body
Uprobe: Additional argumentarch_uprobetouprobe_write_opcode()
Uprobes: Support SDT markers having reference count (semaphore) Uprobes/sdt: Prevent multiple reference counter for same uprobe trace_uprobe/sdt: Prevent multiple reference counter for same uprobe
OK, feel free to add
Quoted 1 line
perf probe: Support SDT markers having reference counter (semaphore)
I didn't even try to read this patch.
Oleg.