You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem? Please describe.
Have fuzzing be an automated task so no one needs to do anything other than fix the bugs
Describe the solution you'd like
get project approved then integrate related infra
Describe alternatives you've considered
manual fuzzing is tedious
Additional context
Fuzzing found a fairly large number of bugs in a short period of time and has prove to be worth the time to integrate.
@ranj063 can you give me a google account email that I can use to add you. @xiulipan if we get approved we will need to start having jenkins uploading builds to a bucket and will probably need your help
So got it to a buildable stage, but this is going to be a challenge. oss-fuzz builds with clang and all our infra is built around gcc. Will need to make some changes.
So, buried deep within the closed issues of oss-fuzz is what I feared. It does not support the default method of fuzzing we are doing. It only supports persistent fuzzing using shims. To continue using the current fuzzing method we will have to do our own ad hoc system. As for oss-fuzz I am working to write a shim that it can work with.
oss-fuzz is going well, patches posted for all fuzzer bugs, few infra bugs to post and seed corpus is still missing. Once those are complete we can begin automated fuzzing.
Activity
cujomalainey commentedon Sep 3, 2020
@ranj063 can you give me a google account email that I can use to add you.
@xiulipan if we get approved we will need to start having jenkins uploading builds to a bucket and will probably need your help
[-][FEATURE] Add fuzzer to oss-fuzz[/-][+][FEATURE] Add sof to oss-fuzz[/+]xiulipan commentedon Sep 4, 2020
@cujomalainey Do you mean we will have a google cloud server or bucket to store build binary?
cujomalainey commentedon Sep 4, 2020
That is the idea from what I'm reading in the docs. CI would build what they need then they would run the fuzzers for us.
cujomalainey commentedon Sep 15, 2020
@ranj063 friendly ping so I can send the PR
ranj063 commentedon Sep 17, 2020
@cujomalainey sorry for the delay. here it is
ranjani.sridharan@intel.corp-partner.google.com
cujomalainey commentedon Sep 21, 2020
@xiulipan I was reading old documentation, CI is not needed
cujomalainey commentedon Sep 22, 2020
Project has been accepted into OSS-fuzz. I am working on a trimmed docker file (removing all the xtensa stuff) and a build script for fuzzing
cujomalainey commentedon Sep 23, 2020
So got it to a buildable stage, but this is going to be a challenge. oss-fuzz builds with clang and all our infra is built around gcc. Will need to make some changes.
cujomalainey commentedon Sep 23, 2020
So, buried deep within the closed issues of oss-fuzz is what I feared. It does not support the default method of fuzzing we are doing. It only supports persistent fuzzing using shims. To continue using the current fuzzing method we will have to do our own ad hoc system. As for oss-fuzz I am working to write a shim that it can work with.
cujomalainey commentedon Oct 28, 2020
oss-fuzz is going well, patches posted for all fuzzer bugs, few infra bugs to post and seed corpus is still missing. Once those are complete we can begin automated fuzzing.
#3562
#3561
#3559
#3558
cujomalainey commentedon Dec 28, 2020
this is done, but fuzzer scope is limited. Need cmake changes to expand it. mainly #3617