| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Logic for interrupts:
- interrupt.S raises spl (thus IF cleared)
- interrupt.S EOI
- interrupt.S calls the handler
- for pure in-kernel handlers, they do whatever they want with IF
cleared.
- when a userland handler is registers, queue_intr masks the irq.
- interrupt.S lowers spl with splx_cli, thus IF still cleared
- iret, that sets IF
- later on, userland acks the IRQ, that unmasks the irq
The key to this change is that all interrupts, including IPIs, are
treated the same way. Eg. the spl level is now raised before an IPI and
restored after. Also, EOI is not needed inside irq_acknowledge.
With this change and the experimental change not to dispatch threads
direct to idle processors in the scheduler, I no longer observe kernel
faults, but an occasional hang does occur.
Message-Id: <20231002033906.124427-1-damien@zamaudio.com>
|
| | |
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
5da1aea7ab3c ("Acknoledge interrupt after handler call") moved the IRQ ack
to after calling the handler because of overflows. But that was because the
interrupts were getting enabled at some point. Now that all spl levels
above 0 just disable interrupts, once we have called spl7 we are safe
until splx_cli is called (and even that doesn't release interrupts, only
the eventual iret will).
And if the handler triggers another IRQ, it will be lost, so we do want
to ack the IRQ before handling it.
|
| | |
|
| |
|
|
| |
Message-Id: <20230805154843.2003098-1-damien@zamaudio.com>
|
| |
|
|
|
|
|
| |
* x86_64/interrupt.S: use 64-bit registers as variables could be
stored at high addresses
* x86_64/locore.S: Likewise
Message-Id: <20230212172818.1511405-3-luca@orpolo.org>
|
| |
|
|
| |
This is not actually used.
|
| |
|
|
|
|
|
|
| |
e2fcf261076b ("interrupt: Fix saving irq/ipl when linux drivers are
disabled") changed the stack layout for interrupt handlers, but missed
updating kdb_kintr's code that tries to mangle it.
This restores the control-alt-d shortcut.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
When Linux drivers are disabled, in hardclock() the linux_timer_intr()
call is dropped, and gcc can tail-recursion-optimize the call to
clock_interrupt(). To do so, it overwrites the hardclock() parameters to
suit the clock_interrupt parameters layout. This however means it thrashes
the backups that the interrupt() function had made of irq/ipl, leading to
mayhem.
interrupt should thus really properly separate its irq/ipl backups from the
interrupt function parameters.
Thanks a lot to Etienne Brateau for the tricky investigation!
|
| |
|
|
|
|
|
|
|
| |
rbx was used to compute the irq index in iunit and ivect arrays,
however it should be preserved by pushing it in to the stack. As a
solution, we use rax instead, which is caller-saved.
Signed-off-by: Luca Dariz <luca@orpolo.org>
Message-Id: <20220205175129.309469-4-luca@orpolo.org>
|
| |
|
|
| |
Message-Id: <20210405115921.184572-2-damien@zamaudio.com>
|
| |
|
|
| |
Message-Id: <20210405052916.174771-4-damien@zamaudio.com>
|
| |
|
|
|
|
|
| |
Use --enable-ncpus=x --enable-apic where x > 1 for SMP+APIC support.
Use neither for no SMP and old PIC support.
Message-Id: <20210404050812.145483-1-damien@zamaudio.com>
|
|
|
* x86_64/interrupt.S: New file.
|
