Print full CSTS value when NVMe device failed to init #1401
Conversation
| @@ -226,7 +226,7 @@ impl<T: DeviceBacking> NvmeDriver<T> { | |||
| )) | |||
| .await | |||
| { | |||
| anyhow::bail!("device is gone"); | |||
| anyhow::bail!("device is gone, csts: {:x}", u32::from(bar0.csts())); | |||
There was a problem hiding this comment.
Should these all have slightly different messages too to distinguish them from one another?
There was a problem hiding this comment.
Yes, that was my concern because I can't distinguish one from another, but I thought printing the full csts value will help figuring it out anyway.
| @@ -329,7 +329,7 @@ impl<T: DeviceBacking> NvmeDriver<T> { | |||
| loop { | |||
| let csts = worker.registers.bar0.csts(); | |||
| if u32::from(csts) == !0 { | |||
| anyhow::bail!("device is gone"); | |||
| anyhow::bail!("device is gone, csts: {:x}", u32::from(csts)); | |||
There was a problem hiding this comment.
nit: store the u32::from result in a var, or just hardcode it since we know what it is thanks to the if
There was a problem hiding this comment.
For this one let me add a variable. I think this code makes an assumption that failed emulated device returns 0xffffffff on read, but I am not sure if that's guaranteed.
| @@ -226,7 +226,7 @@ impl<T: DeviceBacking> NvmeDriver<T> { | |||
| )) | |||
| .await | |||
| { | |||
| anyhow::bail!("device is gone"); | |||
| anyhow::bail!("device is gone, csts: {:x}", u32::from(bar0.csts())); | |||
There was a problem hiding this comment.
I don't think you should read the value again. Either leave this one out (we know reset only fails if csts == -1), or have reset return Result<(), u32>, where the error case is the csts value, and log that.
| @@ -565,7 +566,7 @@ impl<T: DeviceBacking> NvmeDriver<T> { | |||
|
|
|||
| // It is expected the device to be alive when restoring. | |||
| if !bar0.csts().rdy() { | |||
| anyhow::bail!("device is gone"); | |||
| anyhow::bail!("device is gone, csts: {:x}", u32::from(bar0.csts())); | |||
There was a problem hiding this comment.
Don't re-read csts, the value could be different on the next read and that could create confusion.
With the emulated devices likely it will be 0x00000000 or 0xFFFFFFFF but we need to distinguish the cases for further diagnostics.