Re: [PATCH 1/2] docs/livepatch: Add new compiler considerations doc

[Marcos Paulo de Souza]

From: mpdesouza@xxxxxxxx

On   Wed, 2 Sep 2020 15:45:33 +0200 (CEST)   Miroslav Benes <mbenes@xxxxxxx> wrote:

> Hi,
> 
> first, I'm sorry for the late reply. Thanks, Josh, for the reminder.
> 
> CCing Nicolai. Nicolai, could you take a look at the proposed 
> documentation too, please? You have more up-to-date experience.
> 
> On Tue, 21 Jul 2020, Joe Lawrence wrote:
> 
> > +Examples
> > +========
> > +
> > +Interprocedural optimization (IPA)
> > +----------------------------------
> > +
> > +Function inlining is probably the most common compiler optimization that
> > +affects livepatching.  In a simple example, inlining transforms the original
> > +code::
> > +
> > +	foo() { ... [ foo implementation ] ... }
> > +
> > +	bar() { ...  foo() ...  }
> > +
> > +to::
> > +
> > +	bar() { ...  [ foo implementation ] ...  }
> > +
> > +Inlining is comparable to macro expansion, however the compiler may inline
> > +cases which it determines worthwhile (while preserving original call/return
> > +semantics in others) or even partially inline pieces of functions (see cold
> > +functions in GCC function suffixes section below).
> > +
> > +To safely livepatch ``foo()`` from the previous example, all of its callers
> > +need to be taken into consideration.  For those callers that the compiler had
> > +inlined ``foo()``, a livepatch should include a new version of the calling
> > +function such that it:
> > +
> > +  1. Calls a new, patched version of the inlined function, or
> > +  2. Provides an updated version of the caller that contains its own inlined
> > +     and updated version of the inlined function
> 
> I'm afraid the above could cause a confusion...
> 
> "1. Calls a new, patched version of the inlined function, or". The 
> function is not inlined in this case. Would it be more understandable to 
> use function names?
> 
> 1. Calls a new, patched version of function foo(), or
> 2. Provides an updated version of bar() that contains its own inlined and 
>    updated version of foo() (as seen in the example above).
> 
> Not to say that it is again a call of the compiler to decide that, so one 
> usually prepares an updated version of foo() and updated version of bar() 
> calling to it. Updated foo() has to be there for non-inlined cases anyway.
> 
> > +
> > +Other interesting IPA examples include:
> > +
> > +- *IPA-SRA*: removal of unused parameters, replace parameters passed by
> > +  referenced by parameters passed by value.  This optimization basically
> 
> s/referenced/reference/
> 
> > +  violates ABI.
> > +
> > +  .. note::
> > +     GCC changes the name of function.  See GCC function suffixes
> > +     section below.
> > +
> > +- *IPA-CP*: find values passed to functions are constants and then optimizes
> > +  accordingly Several clones of a function are possible if a set is limited.
> 
> "...accordingly. Several..."
> 
> [...]
> 
> > +  	void cdev_put(struct cdev *p)
> > +  	{
> > +  		if (p) {
> > +  			struct module *owner = p->owner;
> > +  			kobject_put(&p->kobj);
> > +  			module_put(owner);
> > +  		}
> > +  	}
> 
> git am complained here about whitespace damage.
> 
> [...]
> 
> > +kgraft-analysis-tool
> > +--------------------
> > +
> > +With the -fdump-ipa-clones flag, GCC will dump IPA clones that were created
> > +by all inter-procedural optimizations in ``<source>.000i.ipa-clones`` files.
> > +
> > +kgraft-analysis-tool pretty-prints those IPA cloning decisions.  The full
> > +list of affected functions provides additional updates that the source-based
> > +livepatch author may need to consider.  For example, for the function
> > +``scatterwalk_unmap()``:
> > +
> > +::
> > +
> > +  $ ./kgraft-ipa-analysis.py --symbol=scatterwalk_unmap aesni-intel_glue.i.000i.ipa-clones
> > +  Function: scatterwalk_unmap/2930 (include/crypto/scatterwalk.h:81:60)
> > +    isra: scatterwalk_unmap.isra.2/3142 (include/crypto/scatterwalk.h:81:60)
> > +      inlining to: helper_rfc4106_decrypt/3007 (arch/x86/crypto/aesni-intel_glue.c:1016:12)
> > +      inlining to: helper_rfc4106_decrypt/3007 (arch/x86/crypto/aesni-intel_glue.c:1016:12)
> > +      inlining to: helper_rfc4106_encrypt/3006 (arch/x86/crypto/aesni-intel_glue.c:939:12)
> > +
> > +    Affected functions: 3
> > +      scatterwalk_unmap.isra.2/3142 (include/crypto/scatterwalk.h:81:60)
> > +      helper_rfc4106_decrypt/3007 (arch/x86/crypto/aesni-intel_glue.c:1016:12)
> > +      helper_rfc4106_encrypt/3006 (arch/x86/crypto/aesni-intel_glue.c:939:12)
> 
> The example for the github is not up-to-date. The tool now expects 
> file_list with *.ipa-clones files and the output is a bit different for 
> the recent kernel.
> 
> $ echo arch/x86/crypto/aesni-intel_glue.c.000i.ipa-clones | kgraft-ipa-analysis.py --symbol=scatterwalk_unmap /dev/stdin
> Parsing file (1/1): arch/x86/crypto/aesni-intel_glue.c.000i.ipa-clones
> Function: scatterwalk_unmap/3935 (./include/crypto/scatterwalk.h:59:20) [REMOVED] [object file: arch/x86/crypto/aesni-intel_glue.c.000i.ipa-clones]
>   isra: scatterwalk_unmap.isra.8/4117 (./include/crypto/scatterwalk.h:59:20) [REMOVED]
>     inlining to: gcmaes_crypt_by_sg/4019 (arch/x86/crypto/aesni-intel_glue.c:682:12) [REMOVED] [edges: 4]
>       constprop: gcmaes_crypt_by_sg.constprop.13/4182 (arch/x86/crypto/aesni-intel_glue.c:682:12)
> 
>   Affected functions: 3
>     scatterwalk_unmap.isra.8/4117 (./include/crypto/scatterwalk.h:59:20) [REMOVED]
>     gcmaes_crypt_by_sg/4019 (arch/x86/crypto/aesni-intel_glue.c:682:12) [REMOVED]
>     gcmaes_crypt_by_sg.constprop.13/4182 (arch/x86/crypto/aesni-intel_glue.c:682:12)
> 
> 
> 
> The rest looks great. Thanks a lot, Joe, for putting it together.

I think that we should start provinding a "Livepatch creation how-to", something
similar, but for now I believe that some documentation is better than no
documentation. This document can evolve to reach such point in the future, but
for now, with Miroslav suggestions applied:

Acked-by: Marcos Paulo de Souza <mpdesouza@xxxxxxxx>

> 
> Miroslav