Re: [RFC PATCH v3] automation: add linker symbol name script

[Nicola Vetrini <nicola.vetrini@bugseng.com>]

On 2024-07-26 00:43, Stefano Stabellini wrote:
> On Thu, 25 Jul 2024, Nicola Vetrini wrote:
>> On 2024-07-25 21:01, victorm.lira@amd.com wrote:
>> > From: Victor Lira <victorm.lira@amd.com>
>> >
>> > Requested-by: Jan Beulich <jbeulich@suse.com>
>> > Signed-off-by: Victor Lira <victorm.lira@amd.com>
>> > ---
>> > Notes:
>> > This is a utilty script for help with the MISRA process.
>> > This script matches all linker symbol names in linker script files for
>> > arm or x86.
>> > Not included are symbol names starting with "." or symbol names enclosed
>> > in quotes since the files dont't use any. The regular expression also does
>> > not match for "&=" and similar compound assignments.
>> > ---
>> > Cc: Jan Beulich <jbeulich@suse.com>
>> > Cc: Stefano Stabellini <sstabellini@kernel.org>
>> > Cc: roberto.bagnara@bugseng.com
>> > Cc: consulting@bugseng.com
>> > Cc: simone.ballarin@bugseng.com
>> > ---
>> > Changes v2:
>> > - address style comments
>> > - updated script to use .lds instead of .lds.S
>> > - remove sample output from patch
>> >
>> > Changes v3:
>> > - use #!/bin/sh
>> > - update error handling and message similar to ../build.sh
>> > ---
>> >  automation/eclair_analysis/linker-symbols.sh | 34 ++++++++++++++++++++
>> >  1 file changed, 34 insertions(+)
>> >  create mode 100755 automation/eclair_analysis/linker-symbols.sh
>> >
>> > diff --git a/automation/eclair_analysis/linker-symbols.sh
>> > b/automation/eclair_analysis/linker-symbols.sh
>> > new file mode 100755
>> > index 0000000000..61790fb281
>> > --- /dev/null
>> > +++ b/automation/eclair_analysis/linker-symbols.sh
>> > @@ -0,0 +1,34 @@
>> > +#!/bin/sh
>> > +
>> > +# Stop immediately if any executed command has exit status different from
>> > 0.
>> > +set -e
>> > +
>> > +# Extract linker symbol names (except those starting with ".") from
>> > assignments.
>> > +
>> > +script_name=$(basename "$0")
>> > +script_dir="$(
>> > +  cd "$(dirname "$0")"
>> > +  echo "${PWD}"
>> > +)"
>> > +src_dir="${script_dir}/../.."
>> > +
>> > +fatal() {
>> > +  echo "${script_name}: $*" >&2
>> > +  exit 1
>> > +}
>> > +
>> > +usage() {
>> > +  fatal "Usage: ${script_name} <arm|x86>"
>> > +}
>> > +
>> > +if [ $# -ne 1 ]; then
>> > +  usage
>> > +fi
>> > +
>> > +filepath="${src_dir}/xen/arch/${1}/xen.lds"
>> > +
>> > +if [ ! -f "$filepath" ]; then
>> > +  fatal "Could not find ${1} linker script. Must be run after arm/x86
>> > build."
>> > +fi
>> > +
>> 
>> A doubt I came across now: since this script must be run after the 
>> build (and
>> hence the analysis), but the configuration must be generated before 
>> the
>> analysis, the only way this could work in my opinion is this:
>> 
>> - a build without analysis is performed, just enough to build xen.lds 
>> (maybe
>> there is a specific Makefile target to do this)
>> - generate the configuration, then clean everything and then run the 
>> analysis
> 
> Yes, that is one option and it should work. The other option is to run
> this script against the xen.lds.S files instead (if it works.)
> 

However, as Jan pointed out in v1, this second option may yield 
incomplete results.
The way we can notice if there's a discrepancy is by running the 
analysis with the automatically generated deviation and tell whether the 
resulting reports for Rule 18.2 there are any about linker-defined 
symbols.

> 
>> > +sed -n "s/^\s*\([a-zA-Z_][a-zA-Z_0-9.\-]*\)\s*=.*;.*$/\1/p" "$filepath"

-- 
Nicola Vetrini, BSc
Software Engineer, BUGSENG srl (https://bugseng.com)