fuzz_nvme_driver: use additional arbitrary inputs #628
Conversation
…y_tag Taking care of all the use-arbitrary-inputs
gurasinghMS
changed the title
mattkur
changed the title
mattkur
changed the title
| .await | ||
| .unwrap(); | ||
|
|
||
| let device = FuzzEmulatedDevice::new(nvme, msi_set, mem); | ||
| let nvme_driver = NvmeDriver::new(&driver_source, 64, device).await.unwrap(); // TODO: [use-arbitrary-input] | ||
| let namespace = nvme_driver.namespace(1).await.unwrap(); // TODO: [use-arbitrary-input] | ||
| let nvme_driver = NvmeDriver::new(&driver_source, arbitrary_data::<u32>()?, device).await.unwrap(); |
There was a problem hiding this comment.
I don't think this will work ... in development of #585 , I tried to use arbitrary cpu here and ran into panic when issuing IO (and in other places).
So ... did your fuzzer not run into panics here? :)
There was a problem hiding this comment.
That's a great point. Local runs of the fuzzer don't seem to panic because of bad cpu (Maybe it is using a different seed to begin with?). Anywho, i will fix this :)
There was a problem hiding this comment.
Or, just wait for my PR and then revisit. This is one of those wonky areas - one of my devices ran into these quite quickly, but others have run for hours without any change.
There was a problem hiding this comment.
(by that, I mean - don't make the cpu arbitrary)
|
Thanks for the PR! Please fix the ci failures and make the cpu count fix, and then I'll take another look. |
Previously there were some hard coded values in the nvme fuzzer. These values are now being determined using the fuzzer input with arbitrary_data::<>()? command. See Issue: #521