LTTng kernel integration roadmap, update
(sorry for repost, I messed up lkml address in the original post, please reply to this one)
Hi everyone,
Let's see what the current status is and what we still have to focus on before submission. First I'd like to say a big thanks to Zhaolei from Fujitsu who did a great job for the debugfs-based LTTng control we now have.
- trace_clock()
- Port to the -tip tree. Submit. Should be ready for inclusion after the few rounds of review on LKML it had.
Assigned-to: Mathieu Desnoyers <[email protected]>
- LTTng instrumentation
- Port to the -tip tree, submit incrementally.
Assigned-to: Mathieu Desnoyers <[email protected]>
- Add periodical subbuffer flush for streaming
- Will be required by ascii text output.
- Also requires modification to lttd so it can only transmit the buffer data, without the padding. Some modification to LTTV might be needed to cope with variable-sized subbuffers (we would need an index to be able to seek with O(log(num subbuffers)) as we currently do, or we would have to use sparse files for trace files).
Assigned-to: François Prenoveau <[email protected]>
- Marker ID management
Background :
The markers are currently used to identify the numerical event IDs associated with a named event as well as the event types (format string) associated with that event. Those two informations are kept in a table written in the metadata channel. A "channel" could be thought as being a set of per-cpu buffers, where information transits. This information would typically be related to a specific tracer. Note, however, that the information generated by a tracer should always be parseable generically and be transformable into text output by a simple parser, because we want to permit analysis across information logged by the various tracer.
Action item :
The current event IDs are global to the whole kernel. It would be beneficial to change their scope so they are per-channel instead, because we can then encode most events in the low-order bits of the event ID bits, therefore keeping events as small as possible.
- Channel management
Background :
LTTng currently has the following statically declared channels :
cpu (default channel)
metadata (marker name, marker ID, marker format string table) interrupts (interrupt handlers table, interrupt handler connexion events)
modules (loaded module table, module load/free events) network (active interfaces table, interface up/down events) processes (processes existing at trace start table, scheduling events)
Action item :
But if we want to integrate ftrace easily within this framework, we would like to offer an API so tracers can dynamically register a new channel whenever they are loaded.
Note that the marker ID management and channel management are closely tied together; IDs would be per-channel, and channels would be registered by tracers.
- trimmed-down lttv for the kernel tree
Need to look at
git://git.kernel.org/pub/scm/linux/kernel/git/sam/test.git#master
tests/ directory, which permits building userspace tools with the Linux kernel. I recently got the idea of populating debugfs with userspace tools that would sits in kernel module data. How (in)sane does this idea look ? That would seems like a rather good solution to ship the userspace tool with the kernel.
- Port ftrace to LTTng
ftrace could ask LTTng to create channels for its data, and would use this handle to record into the buffers. Given that LTTng allows multiple traces to be taken at once, and given LTTng exports all its trace control primitives as in-kernel API, ftrace could use those to automatically control tracing behavior from within the kernel wrt the buffers for which it produces _and_ consumes information.
I think we might need to add the notion of traces for which buffers are meant to be consumed within the kernel (private) and not exported to userspace. Or maybe the current reference counting which insure that at most one reader is connected on a trace buffer, is enough.
- ASCII text output.
- Merge sort done within the kernel. Binary-to-text translation also done within the kernel.
Assigned-to: Zhaolei <[email protected]>
Assigned-to: Lai Jiangshan <[email protected]>
Mathieu
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Quoted 8 lines · Mathieu Desnoyers
- trimmed-down lttv for the kernel tree
Need to look at
git://git.kernel.org/pub/scm/linux/kernel/git/sam/test.git#master
tests/ directory, which permits building userspace tools with the Linux kernel. I recently got the idea of populating debugfs with userspace tools that would sits in kernel module data. How (in)sane does this idea look ? That would seems like a rather good solution to ship the userspace tool with the kernel.
Completely insane I'd say. Debugfs is per defintion populated from kernelspace, and adding another mixed user/kernel populated fs like devfs is a rather bad idea. I don't quite see any point in shipping lttv with the kernel tree either, it shouldn't really need knowledge of the kernel version.
Quoted 10 lines · * Christoph Hellwig ([email protected]) wrote:
Quoted 8 lines · Mathieu Desnoyers
- trimmed-down lttv for the kernel tree
Need to look at
git://git.kernel.org/pub/scm/linux/kernel/git/sam/test.git#master
tests/ directory, which permits building userspace tools with the Linux kernel. I recently got the idea of populating debugfs with userspace tools that would sits in kernel module data. How (in)sane does this idea look ? That would seems like a rather good solution to ship the userspace tool with the kernel.
Quoted 6 lines
Completely insane I'd say. Debugfs is per defintion populated from kernelspace, and adding another mixed user/kernel populated fs like devfs is a rather bad idea. I don't quite see any point in shipping lttv with the kernel tree either, it shouldn't really need knowledge of the kernel version.
The key idea behind this is to answer to Thomas Gleixner concerns, who supports that a tracer should output data in text-format only so it can be used with tools kernel developers have on their system, like "cat".
However, getting data out of the kernel efficiently simply cannot be done with such approach. Therefore, LTTng needs its own userspace tools to splice the data out of the kernel efficiently. Another tool is used to pretty-print the binary data into text.
Then the problem becomes : we have to make the userspace tool easy enough to deploy so even Linus can find and use it. ;)
But indeed, the trace buffers are versioned, so if the format changes between kernel versions, the userspace tools will detect it and the user will know it must update its tools. So it's not really a problem there.
The question that prevails is therefore : should we ship userspace binary with the kernel tree at all ? And if yes, how should the resuting executables be packaged and deployed ? Should it be installed in the system along with kernel modules or should it be populated into a filesystem populated by kernelspace ?
Or is it better to do as we have always done and keep the userspace tools separated from the kernel tree ?
Mathieu
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Quoted 17 lines · Mathieu Desnoyers
The key idea behind this is to answer to Thomas Gleixner concerns, who supports that a tracer should output data in text-format only so it can be used with tools kernel developers have on their system, like "cat".
However, getting data out of the kernel efficiently simply cannot be done with such approach. Therefore, LTTng needs its own userspace tools to splice the data out of the kernel efficiently. Another tool is used to pretty-print the binary data into text.
Then the problem becomes : we have to make the userspace tool easy enough to deploy so even Linus can find and use it. ;)
But indeed, the trace buffers are versioned, so if the format changes between kernel versions, the userspace tools will detect it and the user will know it must update its tools. So it's not really a problem there.
Quoted 8 lines
The question that prevails is therefore : should we ship userspace binary with the kernel tree at all ? And if yes, how should the resuting executables be packaged and deployed ? Should it be installed in the system along with kernel modules or should it be populated into a filesystem populated by kernelspace ?
Or is it better to do as we have always done and keep the userspace tools separated from the kernel tree ?
I say keep the user space tools separate as much as possible.
What about having a meta-data file for all binary files. This meta-data could explain the format that is read. Big endian, little endian, the fields and offsets, the event ids etc. This way we will not need a "version" file, which means absolutely nothing if you do not know what comes with that version. Any tool could look at the meta-data file and figure out what is in the buffers.
-- Steve
Quoted 25 lines · * Steven Rostedt ([email protected]) wrote:
Quoted 23 lines · Mathieu Desnoyers
The key idea behind this is to answer to Thomas Gleixner concerns, who supports that a tracer should output data in text-format only so it can be used with tools kernel developers have on their system, like "cat".
However, getting data out of the kernel efficiently simply cannot be done with such approach. Therefore, LTTng needs its own userspace tools to splice the data out of the kernel efficiently. Another tool is used to pretty-print the binary data into text.
Then the problem becomes : we have to make the userspace tool easy enough to deploy so even Linus can find and use it. ;)
But indeed, the trace buffers are versioned, so if the format changes between kernel versions, the userspace tools will detect it and the user will know it must update its tools. So it's not really a problem there.
The question that prevails is therefore : should we ship userspace binary with the kernel tree at all ? And if yes, how should the resuting executables be packaged and deployed ? Should it be installed in the system along with kernel modules or should it be populated into a filesystem populated by kernelspace ?
Quoted 5 lines
Quoted 2 lines
Or is it better to do as we have always done and keep the userspace tools separated from the kernel tree ?
I say keep the user space tools separate as much as possible.
I'd be in favor of that too. We should just document and package it so it's easy to find.
Quoted 8 lines
What about having a meta-data file for all binary files. This meta-data could explain the format that is read. Big endian, little endian, the fields and offsets, the event ids etc. This way we will not need a "version" file, which means absolutely nothing if you do not know what comes with that version. Any tool could look at the meta-data file and figure out what is in the buffers.
-- Steve
This is exactly what I do in LTTng, modulo the fact that I repeat this information also in other buffer headers, but only use the information located in the metadata buffer header. I duplicated the information to make sure all subbuffer headers looks the same, but I could easily change that.
I would however keep a small subbuffer header with a version number for each subbuffers though, just so the parser can "know" what file this is and what metadata should be expected with it. I think about the poor user who lost its metadata file and wonders what tool could open the other tracefiles he has... without a header containing at least a magic number and a version, those files won't be identified. But we can keep this information as minimalistic as possible.
Thanks for the feedback.
Mathieu
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Quoted 50 lines · * Mathieu Desnoyers ([email protected]) wrote:
Quoted 25 lines · * Steven Rostedt ([email protected]) wrote:
Quoted 23 lines · Mathieu Desnoyers
The key idea behind this is to answer to Thomas Gleixner concerns, who supports that a tracer should output data in text-format only so it can be used with tools kernel developers have on their system, like "cat".
However, getting data out of the kernel efficiently simply cannot be done with such approach. Therefore, LTTng needs its own userspace tools to splice the data out of the kernel efficiently. Another tool is used to pretty-print the binary data into text.
Then the problem becomes : we have to make the userspace tool easy enough to deploy so even Linus can find and use it. ;)
But indeed, the trace buffers are versioned, so if the format changes between kernel versions, the userspace tools will detect it and the user will know it must update its tools. So it's not really a problem there.
The question that prevails is therefore : should we ship userspace binary with the kernel tree at all ? And if yes, how should the resuting executables be packaged and deployed ? Should it be installed in the system along with kernel modules or should it be populated into a filesystem populated by kernelspace ?
Quoted 5 lines
Quoted 2 lines
Or is it better to do as we have always done and keep the userspace tools separated from the kernel tree ?
I say keep the user space tools separate as much as possible.
I'd be in favor of that too. We should just document and package it so it's easy to find.
Quoted 8 lines
What about having a meta-data file for all binary files. This meta-data could explain the format that is read. Big endian, little endian, the fields and offsets, the event ids etc. This way we will not need a "version" file, which means absolutely nothing if you do not know what comes with that version. Any tool could look at the meta-data file and figure out what is in the buffers.
-- Steve
This is exactly what I do in LTTng, modulo the fact that I repeat this information also in other buffer headers, but only use the information located in the metadata buffer header. I duplicated the information to make sure all subbuffer headers looks the same, but I could easily change that.
Quoted 8 lines
I would however keep a small subbuffer header with a version number for each subbuffers though, just so the parser can "know" what file this is and what metadata should be expected with it. I think about the poor user who lost its metadata file and wonders what tool could open the other tracefiles he has... without a header containing at least a magic number and a version, those files won't be identified. But we can keep this information as minimalistic as possible.
I now remember why I duplicate the endianness and the type size information in the subbuffer header.
The main use-case for this is userspace tracing on a mixed 32/64-bits system, where I'd like to be able to write from a 32-bits userspace program directly into their own memory buffers. I would therefore have one "channel" per process, or something like that. The thing is : I have to know the pointer size so I know how the event headers are aligned in the traces. However, with mixed 32/64-bits processes, I would happen to have specific data types for specific channels (note that size_t is also special : it changes depending on the compiling options for 32-bits executables). Therefore, it seems to make sense to export this type of information directly in the subbuffer headers.
I think we can still iterate on the userspace tracing aspects of tracing later on, so we don't have to come up with a "perfect solution" right now. I just want us to keep an open mind for the requirements that will come with its implementation and try to plan a little bit ahead.
Mathieu
Quoted 13 lines
Thanks for the feedback.
Mathieu
Signature · 7 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68 _______________________________________________ ltt-dev mailing list [email protected] http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Quoted 10 lines · On Mon, 24 Nov 2008 11:58:40 -0500 (EST) Steven Rostedt <[email protected]>…
Quoted 8 lines
The question that prevails is therefore : should we ship userspace binary with the kernel tree at all ? And if yes, how should the resuting executables be packaged and deployed ? Should it be installed in the system along with kernel modules or should it be populated into a filesystem populated by kernelspace ?
Or is it better to do as we have always done and keep the userspace tools separated from the kernel tree ?
I say keep the user space tools separate as much as possible.
So... we're now planning to build even more userspace tools into the kernel merely because we don't know how to deliver userspace tools in userspace?
Worse, those kernel-based userspace tools will simply be presenting a textual form of something which the kernel already makes available in a binary form?
This is so lame.
Any proposal to fix all this will result in vast amounts of inconclusive chin-scratching again. Last time this came up, we ended up with the old impractical put-it-in-util-linux chestnut. I think the time has come for us to admit that this isn't working out.
Mathieu, if you're feeling keen I'd suggest that you just type `mkdir -p userspace/lttng' and build your userspace tools in there. One they are there and real, the issues around versioning, installation and delivery will still need tobe sorted out, but we need to start somewhere.
Quoted 3 lines
Mathieu, if you're feeling keen I'd suggest that you just type `mkdir -p userspace/lttng' and build your userspace tools in there.
Maybe this can help...
Sam
usr/Makefile +2 -0
@@ -2,6 +2,8 @@ # kbuild file for usr/ - including initramfs image # +subdir-y := lttng + klibcdirs:; PHONY += klibcdirs
usr/lttng/Makefile new file +4 -0
@@ -0,0 +1,4 @@ + +always := ltt + +hostprogs-y := ltt
usr/lttng/ltt.c new file +7 -0
@@ -0,0 +1,7 @@ +#include <stdio.h> + +int main (int ac,char *argv[]) +{ +printf ("lttng\n" ); +return 0 ; +}
Quoted 14 lines · Sam Ravnborg
Quoted 3 lines
Mathieu, if you're feeling keen I'd suggest that you just type `mkdir -p userspace/lttng' and build your userspace tools in there.
Maybe this can help...
Sam
usr/lttng/Makefile new file +0
-0
usr/lttng/Makefile line 4
Quoted 4 lines
usr/lttng/Makefile new file +4
-0@@ -0,0 +1,4 @@ + +always := ltt + +hostprogs-y := ltt
hostprogs are intended to be run on the host during the compilation process, right? What we really want is something that allows us to build userspace programs intended for use on the target, not the host, since they aren't going to be used when the system is compiled, but when the system is run.
This also brings up some interesting questions such as where would we install these userspace programs (which could be kernel version specific), and how they would be packaged. In some ways, this is similar to a previously unsolved and unresolved problem with firmware which people want to move out of the kernel (but not for legal reasons, they swear, but because most of the time the firmware stays constant from version to version --- except when it doesn't).
The same arguments that are used for ejecting firmware from the kernel and not wanting to ship multiple versions of binaries with the kernel can also be used for these sorts of low-level userspace binaries which may be kernel version specific which are often duplicated.
I happen to disagree with those who wish to eject firmware from the kernel sources (because I don't believe it is for technology reasons, but because they want to be able to chant, "Unclean, Unclean!" as they type rm -rf firmware). And I think we should need to be able to distribute userspace binaries in the kernel, because of the versioning issues and so we can guarantee that we get upgraded userspace tools into the hands of users when they upgrade to the latest kernel.org kernel and need certian updated userspace utilities where we want the stable kernel interface to be located at a kernel-shipped userspace program or library.
However, we do need to understand that there are issues around building target userspace progs, where they get installed, changing the kernel packaging systems to include them in the kernel image package, creating some convention so users can set the appropriate search path in their shells depending on which kernel they have booted, etc. These are not insurmountable problems. But they are ones that we need to solve first.
- Ted
Quoted 9 lines · On Tue, 25 Nov 2008 12:59:37 -0500 Theodore Tso <[email protected]> wrote:
Quoted 8 lines · Sam Ravnborg
Quoted 3 lines
Mathieu, if you're feeling keen I'd suggest that you just type `mkdir -p userspace/lttng' and build your userspace tools in there.
Maybe this can help...
Sam
Quoted 10 lines
usr/lttng/Makefile line 4
Quoted 10 lines
usr/lttng/Makefile new file +4
-0@@ -0,0 +1,4 @@ + +always := ltt + +hostprogs-y := ltt
^^ good start!
Quoted 30 lines
hostprogs are intended to be run on the host during the compilation process, right? What we really want is something that allows us to build userspace programs intended for use on the target, not the host, since they aren't going to be used when the system is compiled, but when the system is run.
This also brings up some interesting questions such as where would we install these userspace programs (which could be kernel version specific), and how they would be packaged. In some ways, this is similar to a previously unsolved and unresolved problem with firmware which people want to move out of the kernel (but not for legal reasons, they swear, but because most of the time the firmware stays constant from version to version --- except when it doesn't).
The same arguments that are used for ejecting firmware from the kernel and not wanting to ship multiple versions of binaries with the kernel can also be used for these sorts of low-level userspace binaries which may be kernel version specific which are often duplicated.
I happen to disagree with those who wish to eject firmware from the kernel sources (because I don't believe it is for technology reasons, but because they want to be able to chant, "Unclean, Unclean!" as they type rm -rf firmware). And I think we should need to be able to distribute userspace binaries in the kernel, because of the versioning issues and so we can guarantee that we get upgraded userspace tools into the hands of users when they upgrade to the latest kernel.org kernel and need certian updated userspace utilities where we want the stable kernel interface to be located at a kernel-shipped userspace program or library.
Quoted 8 lines
However, we do need to understand that there are issues around building target userspace progs, where they get installed, changing the kernel packaging systems to include them in the kernel image package, creating some convention so users can set the appropriate search path in their shells depending on which kernel they have booted, etc. These are not insurmountable problems. But they are ones that we need to solve first.
I think when you add in internationalisation, they _are_ insurmountable problems. I don't think that full-on shipping of userspace-ready applications by the kernel team will prove practical.
But let's look at the problem which we're actually trying to solve. Developer A wishes to write some kernel monitoring/controlling code, so he is forced to stick it on his website, keep reminding people to download updates, act as an independent target of other people's patches, etc, etc. It's all a pain and horror, so developer A gives up and implements his userspace code in the kernel instead. It is, as a result, technically inferior and English-only, but at least it got there.
So I'd propose that userspace/foo/ be viewed as a code *development* location, not a *delivery* location. Once a tool has reached a sufficient level of maturity, stability and usefulness then it's time to ask the util-linux people "please take this", then remove it from the kernel tree. Or to create a separate package altogether, and to remove the original from the kernel tree.
So userspace/foo/ doesn't have to be *fancy*. It's a three- to twelve- month thing where code can sit while kernel developers evolve it. Its main consumers at that time will be kernel developers and testers.
Quoted 9 lines · Theodore Tso
Quoted 8 lines · Sam Ravnborg
Quoted 3 lines
Mathieu, if you're feeling keen I'd suggest that you just type `mkdir -p userspace/lttng' and build your userspace tools in there.
Maybe this can help...
Sam
Quoted 16 lines
usr/lttng/Makefile line 4
Quoted 10 lines
usr/lttng/Makefile new file +4
-0@@ -0,0 +1,4 @@ + +always := ltt + +hostprogs-y := ltthostprogs are intended to be run on the host during the compilation process, right? What we really want is something that allows us to build userspace programs intended for use on the target, not the host, since they aren't going to be used when the system is compiled, but when the system is run.
Yes. And if this fly I will add support for that.
The majority build their kernel on the same host where they run it, or at least a binary compatible host.
Quoted 4 lines
This also brings up some interesting questions such as where would we install these userspace programs (which could be kernel version specific), and how they would be packaged.
We already have similar userspace tools which we have no problem distributing. Think of checkpatch, bloat-o-meter, cleanfile etc.
They happens to be scripts and not binaries. But otherwise I see no big difference.
Sam
Quoted 6 lines · * Sam Ravnborg ([email protected]) wrote:
Quoted 3 lines
Mathieu, if you're feeling keen I'd suggest that you just type `mkdir -p userspace/lttng' and build your userspace tools in there.
Maybe this can help...
Hi Sam,
Thanks for the code snipped, I'll start from there for my kernel-tree integrated userspace tools when I come back from travelling around Europe.
Mathieu
Quoted 44 lines
Sam
usr/Makefile +2
-0@@ -2,6 +2,8 @@ # kbuild file for usr/ - including initramfs image # +subdir-y := lttng + klibcdirs:; PHONY += klibcdirsusr/lttng/Makefile new file +4
-0@@ -0,0 +1,4 @@ + +always := ltt + +hostprogs-y := lttusr/lttng/ltt.c new file +7
-0@@ -0,0 +1,7 @@ +#include <stdio.h> + +int main (int ac,char *argv[]) +{ +printf ("lttng\n" ); +return 0 ; +}_______________________________________________
ltt-dev mailing list
[email protected]
http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Quoted 28 lines · * Andrew Morton ([email protected]) wrote:
Quoted 10 lines · On Mon, 24 Nov 2008 11:58:40 -0500 (EST) Steven Rostedt <[email protected]>…
Quoted 8 lines
The question that prevails is therefore : should we ship userspace binary with the kernel tree at all ? And if yes, how should the resuting executables be packaged and deployed ? Should it be installed in the system along with kernel modules or should it be populated into a filesystem populated by kernelspace ?
Or is it better to do as we have always done and keep the userspace tools separated from the kernel tree ?
I say keep the user space tools separate as much as possible.
So... we're now planning to build even more userspace tools into the kernel merely because we don't know how to deliver userspace tools in userspace?
Worse, those kernel-based userspace tools will simply be presenting a textual form of something which the kernel already makes available in a binary form?
This is so lame.
Any proposal to fix all this will result in vast amounts of inconclusive chin-scratching again. Last time this came up, we ended up with the old impractical put-it-in-util-linux chestnut. I think the time has come for us to admit that this isn't working out.
Quoted 6 lines
Mathieu, if you're feeling keen I'd suggest that you just type `mkdir -p userspace/lttng' and build your userspace tools in there. One they are there and real, the issues around versioning, installation and delivery will still need tobe sorted out, but we need to start somewhere.
Yes, I like the idea of having a kernel-developer oriented version of the userspace tools, which may not have packaging requirements as strong as standard userspace packages. Just keeping them around with the kernel build tree would be a good start, and we could probably go as far as installing them along with the kernel modules in /lib/modules/uname -r.
Mathieu
Quoted 6 lines
_______________________________________________
ltt-dev mailing list
[email protected]
http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Quoted 3 lines · Mathieu Desnoyers
The key idea behind this is to answer to Thomas Gleixner concerns, who supports that a tracer should output data in text-format only so it can be used with tools kernel developers have on their system, like "cat".
Last time I talked to Thomas and others it wasn't "text only" but text also in that you can just cat a debugfs file for simple tracing. And I totally agree with that. If you don't trace too much that's absolutely enough. It's not enough for large traces on highly loaded enterprise or high-end embedded systems or things like flight recorder tracing, but for those requiring an external viewer is just fine.
Quoted 5 lines · * Christoph Hellwig ([email protected]) wrote:
Quoted 3 lines · Mathieu Desnoyers
The key idea behind this is to answer to Thomas Gleixner concerns, who supports that a tracer should output data in text-format only so it can be used with tools kernel developers have on their system, like "cat".
Quoted 7 lines
Last time I talked to Thomas and others it wasn't "text only" but text also in that you can just cat a debugfs file for simple tracing. And I totally agree with that. If you don't trace too much that's absolutely enough. It's not enough for large traces on highly loaded enterprise or high-end embedded systems or things like flight recorder tracing, but for those requiring an external viewer is just fine.
Ok, thanks for the clarification. This will be pretty much straightforward to implement.
Mathieu
Quoted 6 lines
_______________________________________________
ltt-dev mailing list
[email protected]
http://lists.casi.polymtl.ca/cgi-bin/mailman/listinfo/ltt-dev
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Quoted 11 lines · Christoph Hellwig
Quoted 3 lines · Mathieu Desnoyers
The key idea behind this is to answer to Thomas Gleixner concerns, who supports that a tracer should output data in text-format only so it can be used with tools kernel developers have on their system, like "cat".
Last time I talked to Thomas and others it wasn't "text only" but text also in that you can just cat a debugfs file for simple tracing. And I totally agree with that. If you don't trace too much that's absolutely enough. It's not enough for large traces on highly loaded enterprise or high-end embedded systems or things like flight recorder tracing, but for those requiring an external viewer is just fine.
I agree that we should have a text-based output format for kernel hackers, as well as a binary high volume format. I'd argue also that since the text-based one would be used by kernel hackers, we don't need to worry so much about keeping the format absolutely stable from kernel version to kernel version (and it's not worth it to add version tags to the text stream); people who care about a long-term, stable interface should use the binary output format, IMHO.
I'm not sure everyone will agree with me, but I think it helps to simplify the effort behind the text-based output.
- Ted
Quoted 10 lines · Theodore Tso
I agree that we should have a text-based output format for kernel hackers, as well as a binary high volume format. I'd argue also that since the text-based one would be used by kernel hackers, we don't need to worry so much about keeping the format absolutely stable from kernel version to kernel version (and it's not worth it to add version tags to the text stream); people who care about a long-term, stable interface should use the binary output format, IMHO.
I'm not sure everyone will agree with me, but I think it helps to simplify the effort behind the text-based output.
Yes, agreed.
Quoted 13 lines · Mathieu Desnoyers
- Marker ID management
Background :
The markers are currently used to identify the numerical event IDs associated with a named event as well as the event types (format string) associated with that event. Those two informations are kept in a table written in the metadata channel. A "channel" could be thought as being a set of per-cpu buffers, where information transits. This information would typically be related to a specific tracer. Note, however, that the information generated by a tracer should always be parseable generically and be transformable into text output by a simple parser, because we want to permit analysis across information logged by the various tracer.
Quoted 6 lines
Action item :
The current event IDs are global to the whole kernel. It would be beneficial to change their scope so they are per-channel instead, because we can then encode most events in the low-order bits of the event ID bits, therefore keeping events as small as possible.
Hi, Mathieu Desnoyers:
It also required by ASCII text output.
ASCII text output needs get name&fmt from eID, but In current code, I cannot find a reliable way to get name&fmt from eID. And also, in the kernel space, we must get name&fmt from eID as fast as possible by reason of performance. So I cannot use metadata channel for it.
When "Add periodical subbuffer flush for streaming" and "Marker ID management" will done?
Thanks, Lai.