uprobes: Add support to optimize usdt probes on x86_64

[RFC perf/core 03/11] uprobes: Add len argument to uprobe_write_opcode

Posted by Jiri Olsa 2 weeks, 4 days ago

Adding len argument to uprobe_write_opcode as preparation
fo writing longer instructions in following changes.

Signed-off-by: Jiri Olsa <jolsa@kernel.org>
---
 include/linux/uprobes.h |  3 ++-
 kernel/events/uprobes.c | 14 ++++++++------
 2 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
index 28068f9fcdc1..7d23a4fee6f4 100644
--- a/include/linux/uprobes.h
+++ b/include/linux/uprobes.h
@@ -181,7 +181,8 @@ 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 arch_uprobe *auprobe, 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 *insn, int len);
 extern struct uprobe *uprobe_register(struct inode *inode, loff_t offset, loff_t ref_ctr_offset, struct uprobe_consumer *uc);
 extern int uprobe_apply(struct uprobe *uprobe, struct uprobe_consumer *uc, bool);
 extern void uprobe_unregister_nosync(struct uprobe *uprobe, struct uprobe_consumer *uc);
diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
index e9308649bba3..3e275717789b 100644
--- a/kernel/events/uprobes.c
+++ b/kernel/events/uprobes.c
@@ -471,7 +471,7 @@ static int update_ref_ctr(struct uprobe *uprobe, struct mm_struct *mm,
  * Return 0 (success) or a negative errno.
  */
 int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
-			unsigned long vaddr, uprobe_opcode_t opcode)
+			unsigned long vaddr, uprobe_opcode_t *insn, int len)
 {
 	struct uprobe *uprobe;
 	struct page *old_page, *new_page;
@@ -480,7 +480,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
 	bool orig_page_huge = false;
 	unsigned int gup_flags = FOLL_FORCE;
 
-	is_register = is_swbp_insn(&opcode);
+	is_register = is_swbp_insn(insn);
 	uprobe = container_of(auprobe, struct uprobe, arch);
 
 retry:
@@ -491,7 +491,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
 	if (IS_ERR(old_page))
 		return PTR_ERR(old_page);
 
-	ret = verify_opcode(old_page, vaddr, &opcode);
+	ret = verify_opcode(old_page, vaddr, insn);
 	if (ret <= 0)
 		goto put_old;
 
@@ -525,7 +525,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
 
 	__SetPageUptodate(new_page);
 	copy_highpage(new_page, old_page);
-	copy_to_page(new_page, vaddr, &opcode, UPROBE_SWBP_INSN_SIZE);
+	copy_to_page(new_page, vaddr, insn, len);
 
 	if (!is_register) {
 		struct page *orig_page;
@@ -582,7 +582,9 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
  */
 int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr)
 {
-	return uprobe_write_opcode(auprobe, mm, vaddr, UPROBE_SWBP_INSN);
+	uprobe_opcode_t insn = UPROBE_SWBP_INSN;
+
+	return uprobe_write_opcode(auprobe, mm, vaddr, &insn, UPROBE_SWBP_INSN_SIZE);
 }
 
 /**
@@ -598,7 +600,7 @@ int __weak
 set_orig_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr)
 {
 	return uprobe_write_opcode(auprobe, mm, vaddr,
-			*(uprobe_opcode_t *)&auprobe->insn);
+			(uprobe_opcode_t *)&auprobe->insn, UPROBE_SWBP_INSN_SIZE);
 }
 
 /* uprobe should have guaranteed positive refcount */
-- 
2.47.0

Re: [RFC perf/core 03/11] uprobes: Add len argument to uprobe_write_opcode

Posted by Andrii Nakryiko 1 week, 2 days ago

On Tue, Nov 5, 2024 at 5:34 AM Jiri Olsa <jolsa@kernel.org> wrote:
>
> Adding len argument to uprobe_write_opcode as preparation
> fo writing longer instructions in following changes.

typo: for

>
> Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> ---
>  include/linux/uprobes.h |  3 ++-
>  kernel/events/uprobes.c | 14 ++++++++------
>  2 files changed, 10 insertions(+), 7 deletions(-)
>
> diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
> index 28068f9fcdc1..7d23a4fee6f4 100644
> --- a/include/linux/uprobes.h
> +++ b/include/linux/uprobes.h
> @@ -181,7 +181,8 @@ 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 arch_uprobe *auprobe, 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 *insn, int len);

is len in sizeof(uprobe_opcode_t) units or in bytes? it would be good
to make this clearer

it feels like passing `void *` for insns would be better, tbh...



>  extern struct uprobe *uprobe_register(struct inode *inode, loff_t offset, loff_t ref_ctr_offset, struct uprobe_consumer *uc);
>  extern int uprobe_apply(struct uprobe *uprobe, struct uprobe_consumer *uc, bool);
>  extern void uprobe_unregister_nosync(struct uprobe *uprobe, struct uprobe_consumer *uc);
> diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
> index e9308649bba3..3e275717789b 100644
> --- a/kernel/events/uprobes.c
> +++ b/kernel/events/uprobes.c
> @@ -471,7 +471,7 @@ static int update_ref_ctr(struct uprobe *uprobe, struct mm_struct *mm,
>   * Return 0 (success) or a negative errno.
>   */
>  int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
> -                       unsigned long vaddr, uprobe_opcode_t opcode)
> +                       unsigned long vaddr, uprobe_opcode_t *insn, int len)
>  {
>         struct uprobe *uprobe;
>         struct page *old_page, *new_page;
> @@ -480,7 +480,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
>         bool orig_page_huge = false;
>         unsigned int gup_flags = FOLL_FORCE;
>
> -       is_register = is_swbp_insn(&opcode);
> +       is_register = is_swbp_insn(insn);
>         uprobe = container_of(auprobe, struct uprobe, arch);
>
>  retry:
> @@ -491,7 +491,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
>         if (IS_ERR(old_page))
>                 return PTR_ERR(old_page);
>
> -       ret = verify_opcode(old_page, vaddr, &opcode);
> +       ret = verify_opcode(old_page, vaddr, insn);
>         if (ret <= 0)
>                 goto put_old;
>
> @@ -525,7 +525,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
>
>         __SetPageUptodate(new_page);
>         copy_highpage(new_page, old_page);
> -       copy_to_page(new_page, vaddr, &opcode, UPROBE_SWBP_INSN_SIZE);
> +       copy_to_page(new_page, vaddr, insn, len);
>
>         if (!is_register) {
>                 struct page *orig_page;
> @@ -582,7 +582,9 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
>   */
>  int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr)
>  {
> -       return uprobe_write_opcode(auprobe, mm, vaddr, UPROBE_SWBP_INSN);
> +       uprobe_opcode_t insn = UPROBE_SWBP_INSN;
> +
> +       return uprobe_write_opcode(auprobe, mm, vaddr, &insn, UPROBE_SWBP_INSN_SIZE);
>  }
>
>  /**
> @@ -598,7 +600,7 @@ int __weak
>  set_orig_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr)
>  {
>         return uprobe_write_opcode(auprobe, mm, vaddr,
> -                       *(uprobe_opcode_t *)&auprobe->insn);
> +                       (uprobe_opcode_t *)&auprobe->insn, UPROBE_SWBP_INSN_SIZE);
>  }
>
>  /* uprobe should have guaranteed positive refcount */
> --
> 2.47.0
>

Re: [RFC perf/core 03/11] uprobes: Add len argument to uprobe_write_opcode

Posted by Jiri Olsa 1 week ago

On Thu, Nov 14, 2024 at 03:41:03PM -0800, Andrii Nakryiko wrote:
> On Tue, Nov 5, 2024 at 5:34 AM Jiri Olsa <jolsa@kernel.org> wrote:
> >
> > Adding len argument to uprobe_write_opcode as preparation
> > fo writing longer instructions in following changes.
> 
> typo: for
> 
> >
> > Signed-off-by: Jiri Olsa <jolsa@kernel.org>
> > ---
> >  include/linux/uprobes.h |  3 ++-
> >  kernel/events/uprobes.c | 14 ++++++++------
> >  2 files changed, 10 insertions(+), 7 deletions(-)
> >
> > diff --git a/include/linux/uprobes.h b/include/linux/uprobes.h
> > index 28068f9fcdc1..7d23a4fee6f4 100644
> > --- a/include/linux/uprobes.h
> > +++ b/include/linux/uprobes.h
> > @@ -181,7 +181,8 @@ 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 arch_uprobe *auprobe, 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 *insn, int len);
> 
> is len in sizeof(uprobe_opcode_t) units or in bytes? it would be good
> to make this clearer
> 
> it feels like passing `void *` for insns would be better, tbh...

good catch, it's meant as bytes, uprobe_opcode_t is u8 on x86, but other archs
treat it differently, void * might solve that in this function, will check

thanks,
jirka

> 
> 
> 
> >  extern struct uprobe *uprobe_register(struct inode *inode, loff_t offset, loff_t ref_ctr_offset, struct uprobe_consumer *uc);
> >  extern int uprobe_apply(struct uprobe *uprobe, struct uprobe_consumer *uc, bool);
> >  extern void uprobe_unregister_nosync(struct uprobe *uprobe, struct uprobe_consumer *uc);
> > diff --git a/kernel/events/uprobes.c b/kernel/events/uprobes.c
> > index e9308649bba3..3e275717789b 100644
> > --- a/kernel/events/uprobes.c
> > +++ b/kernel/events/uprobes.c
> > @@ -471,7 +471,7 @@ static int update_ref_ctr(struct uprobe *uprobe, struct mm_struct *mm,
> >   * Return 0 (success) or a negative errno.
> >   */
> >  int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
> > -                       unsigned long vaddr, uprobe_opcode_t opcode)
> > +                       unsigned long vaddr, uprobe_opcode_t *insn, int len)
> >  {
> >         struct uprobe *uprobe;
> >         struct page *old_page, *new_page;
> > @@ -480,7 +480,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
> >         bool orig_page_huge = false;
> >         unsigned int gup_flags = FOLL_FORCE;
> >
> > -       is_register = is_swbp_insn(&opcode);
> > +       is_register = is_swbp_insn(insn);
> >         uprobe = container_of(auprobe, struct uprobe, arch);
> >
> >  retry:
> > @@ -491,7 +491,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
> >         if (IS_ERR(old_page))
> >                 return PTR_ERR(old_page);
> >
> > -       ret = verify_opcode(old_page, vaddr, &opcode);
> > +       ret = verify_opcode(old_page, vaddr, insn);
> >         if (ret <= 0)
> >                 goto put_old;
> >
> > @@ -525,7 +525,7 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
> >
> >         __SetPageUptodate(new_page);
> >         copy_highpage(new_page, old_page);
> > -       copy_to_page(new_page, vaddr, &opcode, UPROBE_SWBP_INSN_SIZE);
> > +       copy_to_page(new_page, vaddr, insn, len);
> >
> >         if (!is_register) {
> >                 struct page *orig_page;
> > @@ -582,7 +582,9 @@ int uprobe_write_opcode(struct arch_uprobe *auprobe, struct mm_struct *mm,
> >   */
> >  int __weak set_swbp(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr)
> >  {
> > -       return uprobe_write_opcode(auprobe, mm, vaddr, UPROBE_SWBP_INSN);
> > +       uprobe_opcode_t insn = UPROBE_SWBP_INSN;
> > +
> > +       return uprobe_write_opcode(auprobe, mm, vaddr, &insn, UPROBE_SWBP_INSN_SIZE);
> >  }
> >
> >  /**
> > @@ -598,7 +600,7 @@ int __weak
> >  set_orig_insn(struct arch_uprobe *auprobe, struct mm_struct *mm, unsigned long vaddr)
> >  {
> >         return uprobe_write_opcode(auprobe, mm, vaddr,
> > -                       *(uprobe_opcode_t *)&auprobe->insn);
> > +                       (uprobe_opcode_t *)&auprobe->insn, UPROBE_SWBP_INSN_SIZE);
> >  }
> >
> >  /* uprobe should have guaranteed positive refcount */
> > --
> > 2.47.0
> >