Skip to contentSkip to conversation

Understanding XIP, XPSR, XFS registers

DISCUSSION Keith Owens 3 participants 3 messages Aug 19, 2005

lore
Share & tools
lore
#0 lore raw

Resuming after MCA/INIT when psr.ic = 1 is relatively easy and it appears to work with the new MCA/INIT handlers. Resuming after MCA/INIT when psr.ic = 0 (i.e. the event occurred in an interrupt handler) is proving to be a problem.

When MCA/INIT occurs and psr.ic = 0, what is in X{IP,PSR,FS} in PAL minstate? Is it the original values of I{IP,PSR,FS}, with I{IP,PSR,FS} replaced by the values when MCA/INIT was delivered? Or is I{IP,PSR,FS} left alone (i.e. they still contain the values delivered to the interrupt handler), with X{IP,PSR,FS} set to the values when MCA/INIT was delivered?

Common sense says that I{IP,PSR,FS} is left alone when psr.ic = 0, with the new values being stored in X{IP,PSR,FS}. However Intel Itanium Architecture Software Developer's Manual Volume 2: System Architecture (24531804.pdf), 11.3.3 Returning to the Interrupted Process, implies the reverse.

#1 lore raw

My understanding is that X{IP,PSR,FS} contain the old values of I{IP,PSR,FS}, and I{IP,PSR,FS} contain the real IP,PSR,CFM. I've always found this counterintuitive too, though it might simplify the silicon.

Matt

Quoted 22 lines · Keith Owens

Resuming after MCA/INIT when psr.ic = 1 is relatively easy and it appears to work with the new MCA/INIT handlers. Resuming after MCA/INIT when psr.ic = 0 (i.e. the event occurred in an interrupt handler) is proving to be a problem.

When MCA/INIT occurs and psr.ic = 0, what is in X{IP,PSR,FS} in PAL minstate? Is it the original values of I{IP,PSR,FS}, with I{IP,PSR,FS} replaced by the values when MCA/INIT was delivered? Or is I{IP,PSR,FS} left alone (i.e. they still contain the values delivered to the interrupt handler), with X{IP,PSR,FS} set to the values when MCA/INIT was delivered?

Common sense says that I{IP,PSR,FS} is left alone when psr.ic = 0, with the new values being stored in X{IP,PSR,FS}. However Intel Itanium Architecture Software Developer's Manual Volume 2: System Architecture (24531804.pdf), 11.3.3 Returning to the Interrupted Process, implies the reverse.

-
To unsubscribe from this list: send the line "unsubscribe linux-ia64" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw
In reply to Keith Owens

The manual is correct. You can see that in the (old) MCA backtrace code where we pick up the state from the X registers if PSR.ic=0. I verified at some point that this works as well as it can.

--david

Quoted 23 lines · Keith Owens

Resuming after MCA/INIT when psr.ic = 1 is relatively easy and it appears to work with the new MCA/INIT handlers. Resuming after MCA/INIT when psr.ic = 0 (i.e. the event occurred in an interrupt handler) is proving to be a problem.

When MCA/INIT occurs and psr.ic = 0, what is in X{IP,PSR,FS} in PAL minstate? Is it the original values of I{IP,PSR,FS}, with I{IP,PSR,FS} replaced by the values when MCA/INIT was delivered? Or is I{IP,PSR,FS} left alone (i.e. they still contain the values delivered to the interrupt handler), with X{IP,PSR,FS} set to the values when MCA/INIT was delivered?

Common sense says that I{IP,PSR,FS} is left alone when psr.ic = 0, with the new values being stored in X{IP,PSR,FS}. However Intel Itanium Architecture Software Developer's Manual Volume 2: System Architecture (24531804.pdf), 11.3.3 Returning to the Interrupted Process, implies the reverse.

-
To unsubscribe from this list: send the line "unsubscribe linux-ia64" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

Signature · 3 lines
Mosberger Consulting LLC, voice/fax: 510-744-9372,
http://www.mosberger-consulting.com/
35706 Runckel Lane, Fremont, CA 94536