Date: Mon, 10 May 2004 10:32:04 +0200 From: Joerg Wunsch <j@ida.interface-business.de> To: sparc64@freebsd.org Cc: Thomas Moestl <t.moestl@tu-bs.de> Subject: Re: ebus resource allocation error? Message-ID: <20040510103204.B77963@ida.interface-business.de> In-Reply-To: <20040509213455.GA744@timesink.dyndns.org>; from t.moestl@tu-bs.de on Sun, May 09, 2004 at 11:34:55PM %2B0200 References: <20040509171444.B63877@ida.interface-business.de> <20040509213455.GA744@timesink.dyndns.org>
next in thread | previous in thread | raw e-mail | index | archive | help
As Thomas Moestl wrote: > From e450 firmware property dumps I have lying around it seems > likely that 0x7e5 is the power failure notification interrupt of one > of the Psycho PCI bridges. The Psycho driver does not allocate it > with RF_SHAREABLE (and neither does your code), thus the error. Hmm, now that you mention it, it seems OFW indeed assigns this IRQ to two devices: $ /sbin/dmesg|fgrep 2021 ebus0: <power> addr 0x1400724000-0x1400724003 irq 2034,2021 (no driver attached) ebus0: <SUNW,envctrl> addr 0x1400600000-0x1400600003 irq 2021,2024 (no driver attached) However, due to the `no driver attached' message, I wouldn't even have guessed that the psycho code attached a vector to it. What's the second <power> IRQ for, maybe it indicates a failed power supply? Wouldn't it make more sense to have a separate power driver attaching to it? (The IO regs assigned to <power> belong to the power auxio bits described in the PCIO docs.) > Would an envctrl driver actually need this interrupt, or does this > just indicate that envctrl is the device that really triggers power > failure interrupts? I have no idea about it yet. The SUNW,envctrl hardware supposedly uses a Philips PCF8584 to drive the I²C bus, which can indeed trigger an interrupt. I didn't have a clue at all why there are two IRQs. Most likely the envctrl driver would have no use for the powerfail interrupt, but except for implied knowledge, how would I know which is the `correct' IRQ to attach to? Btw., how do I dump the OFW properties, other than by manually descending through the entire /devices hiearchy? I guess it could be done using a FORTH command, but my FORTH knowledge has been seriously rotten away. ;-) -- J"org Wunsch Unix support engineer joerg_wunsch@interface-systems.de http://www.interface-systems.de/~j/
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20040510103204.B77963>