Skip to contentSkip to conversation

tracing/ftrace: ftrace_bprintk

PATCH In review Frederic Weisbecker 3 participants 5 messages Feb 24, 2009

lore
Share & tools
lore
#0 lore raw

cover tracing/ftrace: ftrace_bprintk

Hi,

These three patches are part of a patchset posted by Lai Jiangshan in december 2008. They introduce a binary version of ftrace_printk() called ftrace_bprintk()

While having the same goal: print a generic message entry into the ring buffer, their approaches are very different.

- ftrace_printk() does the formatting job on tracing time, insert the whole resulting string into the ring buffer, and then the string is printed on output time without a lot of modifications.

- ftrace_bprintk() does no formatting on tracing time. Instead, it looks at the format string to find the types and the numbers of the arguments and directly stores them as-is into the ring-buffer. Then the format string is stored into the ring-buffer too, but only by its address, it is not copied. Then on output time only, the final string is formatted and sent to the user. This gives a result about as fast as a traditional tracer with fixed fields types, except that we can print random types and numbers of fields here.

The first patch adds the generic support for binary formatting. The second adds the support for binary print types on ftrace and the last introduces ftrace_bprintk() which supports safely the modules by listening on the module loading/unloading notifier to keep track of unwanted freed format strings.

Lai Jiangshan (3):
add binary printf
ftrace: infrastructure for supporting binary record
ftrace: add ftrace_bprintk()

 include/asm-generic/vmlinux.lds.h |    3 +
 include/linux/ftrace.h            |   25 ++
 include/linux/module.h            |    5 +
 include/linux/string.h            |    7 +
 kernel/module.c                   |    6 +
 kernel/trace/Kconfig              |    6 +
 kernel/trace/Makefile             |    1 +
 kernel/trace/trace.c              |   70 ++++++
 kernel/trace/trace.h              |   12 +
 kernel/trace/trace_bprintk.c      |  217 ++++++++++++++++++
 kernel/trace/trace_output.c       |   76 +++++++
 lib/Kconfig                       |    3 +
 lib/vsprintf.c                    |  442 +++++++++++++++++++++++++++++++++++++
#1 lore raw
Quoted 25 lines · Frederic Weisbecker

Hi,

These three patches are part of a patchset posted by Lai Jiangshan in december 2008. They introduce a binary version of ftrace_printk() called ftrace_bprintk()

While having the same goal: print a generic message entry into the ring buffer, their approaches are very different.

- ftrace_printk() does the formatting job on tracing time, insert the whole resulting string into the ring buffer, and then the string is printed on output time without a lot of modifications.

- ftrace_bprintk() does no formatting on tracing time. Instead, it looks at the format string to find the types and the numbers of the arguments and directly stores them as-is into the ring-buffer. Then the format string is stored into the ring-buffer too, but only by its address, it is not copied. Then on output time only, the final string is formatted and sent to the user. This gives a result about as fast as a traditional tracer with fixed fields types, except that we can print random types and numbers of fields here.

The first patch adds the generic support for binary formatting. The second adds the support for binary print types on ftrace and the last introduces ftrace_bprintk() which supports safely the modules by listening on the module loading/unloading notifier to keep track of unwanted freed format strings.

Quoted 5 lines

Lai Jiangshan (3):
add binary printf
ftrace: infrastructure for supporting binary record
ftrace: add ftrace_bprintk()

hi,

this seems like a really valuable feature....I'm just wondering about a couple of things....

If the 'brpintk tracer' in trace/trace_bprintk.c is just being used to set an enabled flag for printing out these binary records, then are we better off with just an option flag in the 'trace_options' file?

Second, can we somehow combine ftrace_printk() and ftrace_bprintk(), so that a developer can just use one interface? Perhaps, ftrace_printk calls ftrace_bprintk if binary option flag is set, otherwise, it just outputs things normally.

thanks,

-Jason

#2 lore raw
Quoted 42 lines · Jason Baron
Quoted 25 lines · Frederic Weisbecker

Hi,

These three patches are part of a patchset posted by Lai Jiangshan in december 2008. They introduce a binary version of ftrace_printk() called ftrace_bprintk()

While having the same goal: print a generic message entry into the ring buffer, their approaches are very different.

- ftrace_printk() does the formatting job on tracing time, insert the whole resulting string into the ring buffer, and then the string is printed on output time without a lot of modifications.

- ftrace_bprintk() does no formatting on tracing time. Instead, it looks at the format string to find the types and the numbers of the arguments and directly stores them as-is into the ring-buffer. Then the format string is stored into the ring-buffer too, but only by its address, it is not copied. Then on output time only, the final string is formatted and sent to the user. This gives a result about as fast as a traditional tracer with fixed fields types, except that we can print random types and numbers of fields here.

The first patch adds the generic support for binary formatting. The second adds the support for binary print types on ftrace and the last introduces ftrace_bprintk() which supports safely the modules by listening on the module loading/unloading notifier to keep track of unwanted freed format strings.

Quoted 5 lines

Lai Jiangshan (3):
add binary printf
ftrace: infrastructure for supporting binary record
ftrace: add ftrace_bprintk()

hi,

this seems like a really valuable feature....I'm just wondering about a couple of things....

If the 'brpintk tracer' in trace/trace_bprintk.c is just being used to set an enabled flag for printing out these binary records, then are we better off with just an option flag in the 'trace_options' file?

Quoted 5 lines

Second, can we somehow combine ftrace_printk() and ftrace_bprintk(), so that a developer can just use one interface? Perhaps, ftrace_printk calls ftrace_bprintk if binary option flag is set, otherwise, it just outputs things normally.

Well, ftrace_bprintk() seems to be a worthy and transparent replacement for ftrace_printk() to me. I.e. lets just use this as the new implementation for ftrace_printk().

Would there be any downsides of doing so? I dont see any.

Ingo

#3 lore raw
Quoted 51 lines · Ingo Molnar
Quoted 42 lines · Jason Baron
Quoted 25 lines · Frederic Weisbecker

Hi,

These three patches are part of a patchset posted by Lai Jiangshan in december 2008. They introduce a binary version of ftrace_printk() called ftrace_bprintk()

While having the same goal: print a generic message entry into the ring buffer, their approaches are very different.

- ftrace_printk() does the formatting job on tracing time, insert the whole resulting string into the ring buffer, and then the string is printed on output time without a lot of modifications.

- ftrace_bprintk() does no formatting on tracing time. Instead, it looks at the format string to find the types and the numbers of the arguments and directly stores them as-is into the ring-buffer. Then the format string is stored into the ring-buffer too, but only by its address, it is not copied. Then on output time only, the final string is formatted and sent to the user. This gives a result about as fast as a traditional tracer with fixed fields types, except that we can print random types and numbers of fields here.

The first patch adds the generic support for binary formatting. The second adds the support for binary print types on ftrace and the last introduces ftrace_bprintk() which supports safely the modules by listening on the module loading/unloading notifier to keep track of unwanted freed format strings.

Quoted 5 lines

Lai Jiangshan (3):
add binary printf
ftrace: infrastructure for supporting binary record
ftrace: add ftrace_bprintk()

hi,

this seems like a really valuable feature....I'm just wondering about a couple of things....

If the 'brpintk tracer' in trace/trace_bprintk.c is just being used to set an enabled flag for printing out these binary records, then are we better off with just an option flag in the 'trace_options' file?

Quoted 5 lines

Second, can we somehow combine ftrace_printk() and ftrace_bprintk(), so that a developer can just use one interface? Perhaps, ftrace_printk calls ftrace_bprintk if binary option flag is set, otherwise, it just outputs things normally.

Quoted 7 lines

Well, ftrace_bprintk() seems to be a worthy and transparent replacement for ftrace_printk() to me. I.e. lets just use this as the new implementation for ftrace_printk().

Would there be any downsides of doing so? I dont see any.

Ingo

That's what I think too.
For now I posted Lai's patchset, but my purpose is to replace the old ftrace_printk by this new one. Once it is merged on -tip I'll start iterating around to improve it and replace ftrace_printk.

Quoted 5 lines

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/

#4 lore raw
In reply to Jason Baron
Quoted 32 lines · Jason Baron
Quoted 25 lines · Frederic Weisbecker

Hi,

These three patches are part of a patchset posted by Lai Jiangshan in december 2008. They introduce a binary version of ftrace_printk() called ftrace_bprintk()

While having the same goal: print a generic message entry into the ring buffer, their approaches are very different.

- ftrace_printk() does the formatting job on tracing time, insert the whole resulting string into the ring buffer, and then the string is printed on output time without a lot of modifications.

- ftrace_bprintk() does no formatting on tracing time. Instead, it looks at the format string to find the types and the numbers of the arguments and directly stores them as-is into the ring-buffer. Then the format string is stored into the ring-buffer too, but only by its address, it is not copied. Then on output time only, the final string is formatted and sent to the user. This gives a result about as fast as a traditional tracer with fixed fields types, except that we can print random types and numbers of fields here.

The first patch adds the generic support for binary formatting. The second adds the support for binary print types on ftrace and the last introduces ftrace_bprintk() which supports safely the modules by listening on the module loading/unloading notifier to keep track of unwanted freed format strings.

Quoted 5 lines

Lai Jiangshan (3):
add binary printf
ftrace: infrastructure for supporting binary record
ftrace: add ftrace_bprintk()

Quoted 8 lines

hi,

this seems like a really valuable feature....I'm just wondering about a couple of things....

If the 'brpintk tracer' in trace/trace_bprintk.c is just being used to set an enabled flag for printing out these binary records, then are we better off with just an option flag in the 'trace_options' file?

This tracer does a bit more, it keeps track of the string formats used by a module which uses ftrace_bprintk().
So for now, the use of this facility can't be decoupled from its tracer. In fact, I will try to move this module safety into tracing core so that any tracer will be able to use it.

Quoted 18 lines

Second, can we somehow combine ftrace_printk() and ftrace_bprintk(), so that a developer can just use one interface? Perhaps, ftrace_printk calls ftrace_bprintk if binary option flag is set, otherwise, it just outputs things normally.

thanks,

-Jason