IRC channel logs


back to list of logs

***Server sets mode: +nt
<damo22>youpi: yes i booted native x200 and shelled into the box with ssh using the remove-picmask patch
<damo22>i will rebuild and confirm that
<damo22>but EOI must be getting delivered anyway because the interrupt() entrypoint handles that and calls a handler
<damo22>youpi: mask/unmask have been set to empty
<damo22>so it doesnt even call that
<damo22>ive added a third patch to attempt fixing netdde, testing now
<damo22>although its not broken on x200
<gnu_srs2>damo22: Are you using rump drivers yet?
<damo22>not yet :(
<damo22>im getting a hang on mounting /
<damo22>and ive noticed netdde is hanging on qemu
<damo22>but netdde works fine on x200 native hw
<gnu_srs2>TTT->Things Take Time, good luck ;)
<damo22>i thought i had a fix for netdde on qemu, its still broken though
<damo22>interesting, a similar hang occurs as netdde on qemu after "returning from establish" in rumpdisk
<damo22>but it does not hang there on native hw
<gnu_srs2>a qemu bug then?
<damo22>no youpi solved it i think see ml
<youpi>it looked to me like a level-trigerred vs edge-triggered concern
<youpi>which we need to sort out anyway
<damo22>i think masking all the interrupts and unmasking them at the EOI solves triggering concerns
<youpi>it didn't for me
<damo22>hmm ok
<youpi>I don't see why it would actually
<youpi>masking is just to disable actually interrupting the cpu
<damo22>because the interrupt level gets reset
<youpi>I don't think it can cancel the origin of the trigger
<damo22>so it has to retrigger it
<youpi>that looks like a very odd way to reset it
<damo22>yes, but i think its the only way to do that on a 8259
<youpi>at the very least, it should probably do that only for the concerned irq
<damo22>yes but then we need to do special EOIs not non-specific EOIs
<damo22>but masking a whole pic is one outb
*youpi afk
<damo22>youpi: does master on savannah boot?
<damo22>i noticed youre running inthandlers at spl7, therefore an interrupt cannot interrupt another interrupt i guess
<jrtc27>if it's the PIT->PIC interrupt line, waiting long enough makes the interrupt source go away :)
<damo22>oh wait i am unmasking all the interrupts when you put the masking code back
<damo22>i need to restore the correct mask
<damo22>oops i removed the ACK interrupt code
<damo22>i thought EOI was ack
<damo22>no i didnt, im dreaming
<damo22>youpi: i fixed the hang you were experiencing
<damo22>i keep getting intnull(7) printed repeatedly on real hw now
<youpi>damo22: ah, maybe that's why prtnull was used
<youpi>perhaps we should make the warning printed only once per irq
***Shentino_ is now known as Shentino
***Glider_IRC__ is now known as Glider_IRC