Q: Can I get errors for inconsistencies between trace and binary?

[albertschulz]

Hello :)

Let's imagine the decoder expects a atom for a conditional branch, but gets an atom + an address (as for a indirect branch).
Is it possible to get errors about unexpected trace packets when reconstructing the addresses with OpenCSD?

I would highly appreciate your help. Thanks for this great decoder implementation by the way.

[albertschulz]

Additionally, is there a way to add context-specific memory images?

[mikel-armbb]

Hi, The decoder will only produce errors when it fails to get addresses it does expect. There is nothing in the protocol that prohibits an implementation from outputting addresses at any time, and some ETM implementations do have a "branch broadcast" feature that emits addresses for all branches Regards Mike

…

On Wed, 27 Jul 2022 at 13:31, Albert Schulz ***@***.***> wrote: Hello :) Let's imagine the decoder expects a atom for a conditional branch, but gets an atom + an address (as for a indirect branch). Is it possible to get errors about unexpected trace packets when reconstructing the addresses with OpenCSD? I would highly appreciate your help. Thanks for this great decoder implementation by the way. — Reply to this email directly, view it on GitHub <#45>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADF7BM4PULMBOW3MPBIMUD3VWET3NANCNFSM54ZNYSPQ> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Mike Leach Principal Engineer, ARM Ltd. Manchester Design Centre. UK

[mikel-armbb]

Can you be more specific about the context? There is code in the decode to associate memory images with security / EL states, but nothing beyond that. Mike

…

On Thu, 28 Jul 2022 at 09:41, Albert Schulz ***@***.***> wrote: Additionally, is there a way to add context-specific memory images? — Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Mike Leach Principal Engineer, ARM Ltd. Manchester Design Centre. UK

[albertschulz]

Thanks for your answers.

About the context-specific images:
When tracing linux processes, they share the same address space. So its different how process A with context ID A and process B with context ID B resolves an atom. Depending on the context id, the associated memory image is different. Is there a way to change the memory image on a context id trace message?

[mikel-armbb]

Hi, What you are looking for here is beyond the scope of the library itself. Clients such as perf acheive this by setting a memory image callback to the library and monitoring the packets coming out of the library. If a context ID is switched, then the perf detects this and when the next memory request comes in, it will use the appropriate memory image for that context. Regards Mike

…

On Fri, 29 Jul 2022 at 13:51, Albert Schulz ***@***.***> wrote: Thanks for your answers. About the context-specific images: When tracing linux processes, they share the same address space. So its different how process A with context ID A and process B with context ID B resolves an atom. Depending on the context id, the associated memory image is different. Is there a way to change the memory image on a context id trace message? — Reply to this email directly, view it on GitHub <#45 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ADF7BM3YK3TLJLPMMHSMQ4TVWPHVFANCNFSM54ZNYSPQ> . You are receiving this because you commented.Message ID: ***@***.***>

-- Mike Leach Principal Engineer, ARM Ltd. Manchester Design Centre. UK