From owner-freebsd-mobile Sun Dec 14 22:03:15 1997 Return-Path: Received: (from root@localhost) by hub.freebsd.org (8.8.7/8.8.7) id WAA23665 for mobile-outgoing; Sun, 14 Dec 1997 22:03:15 -0800 (PST) (envelope-from owner-freebsd-mobile) Received: from afs.ntc.mita.keio.ac.jp (afs.ntc.mita.keio.ac.jp [131.113.212.3]) by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id WAA23468 for ; Sun, 14 Dec 1997 22:00:53 -0800 (PST) (envelope-from hosokawa@ntc.keio.ac.jp) Received: (from hosokawa@localhost) by afs.ntc.mita.keio.ac.jp (8.8.8+2.7Wbeta7/3.6Wbeta6-ntc_mailserver1.03) id OAA02926; Mon, 15 Dec 1997 14:57:34 +0900 (JST) Date: Mon, 15 Dec 1997 14:57:34 +0900 (JST) Message-Id: <199712150557.OAA02926@afs.ntc.mita.keio.ac.jp> To: nate@mt.sri.com Cc: mobile@freebsd.org, hosokawa@ntc.keio.ac.jp Subject: Re: Multifunction PC-card support In-Reply-To: Your message of "Sun, 14 Dec 1997 21:17:57 -0700". <199712150417.VAA03645@mt.sri.com> From: hosokawa@ntc.keio.ac.jp (HOSOKAWA Tatsumi) X-Mailer: mnews [version 1.20] 1996-12/08(Sun) Sender: owner-freebsd-mobile@freebsd.org X-Loop: FreeBSD.org Precedence: bulk In article <199712150417.VAA03645@mt.sri.com> nate@mt.sri.com writes: >> > I started the implementation of multifunction PC-card support on >> > FreeBSD based on extended PC-card specification. >> >> I'd be very intersted in knowing how you are doing that, and how you >> plan on changing things in /sys/pccard. I'm sure Mike is as well, since >> this is functionality FreeBSD sorely needs. I only started reading CIS tupples of Mariner card. But unfortunately pccardc dumpcis says that this card has following CIS tupples. -------------------------------------------------------------------------- Configuration data for card in slot 0 Tuple #1, code = 0x1 (Common memory descriptor), length = 2 000: 00 ff Common memory device information: Device number 1, type No device, WPS = OFF Speed = No speed, Memory block size = reserved, 32 units Tuple #2, code = 0x15 (Version 1 info), length = 56 000: 04 01 4d 6f 74 6f 72 6f 6c 61 2c 20 49 6e 63 2e 010: 00 4d 41 52 49 4e 45 52 20 4d 4f 44 45 4d 2f 46 020: 41 58 2f 4c 41 4e 00 56 65 72 2e 20 31 2e 32 2e 030: 30 34 30 38 39 36 00 ff Version = 4.1, Manuf = [Motorola, Inc.],card vers = [MARINER MODEM/FAX/LAN] Addit. info = [Ver. 1.2.040896],[] Tuple #3, code = 0x20 (Manufacturer ID), length = 4 000: 09 01 01 05 PCMCIA ID = 0x109, OEM ID = 0x501 Tuple #4, code = 0x21 (Functional ID), length = 2 000: 02 00 Serial port/modem Tuple #5, code = 0x22 (Functional EXT), length = 4 000: 00 02 0f 1c Serial interface extension: 16550 UART, Parity - Space,Mark,Odd,Even, Tuple #6, code = 0x22 (Functional EXT), length = 11 000: 01 1f 1e 00 04 00 00 00 04 00 00 Modem interface capabilities: Tuple #7, code = 0x22 (Functional EXT), length = 14 000: 02 06 00 3f 1c 02 03 1f 06 00 00 b5 20 00 Data modem services available: Tuple #8, code = 0x22 (Functional EXT), length = 9 000: 06 1f 1e 00 04 00 00 04 00 Modem interface capabilities: Tuple #9, code = 0x22 (Functional EXT), length = 10 000: 13 06 00 0f 00 00 00 b5 20 00 Tuple #10, code = 0x22 (Functional EXT), length = 10 000: 23 06 00 0f 00 00 00 b5 20 00 Tuple #11, code = 0x1a (Configuration map), length = 5 000: 01 45 20 80 77 Reg len = 2, config register addr = 0x8020, last config = 0x45 Registers: XXX-XXX- Tuple #12, code = 0x1b (Configuration entry), length = 18 000: c5 01 99 71 55 1e 4e 23 aa 60 f8 03 07 30 bc 9e 010: 28 00 Config index = 0x5(default) Interface byte = 0x1 (I/O) Vcc pwr: Nominal operating supply voltage: 5 x 1V Max current average over 1 second: 1.5 x 100mA Max current average over 10 ms: 4.5 x 100mA Power down supply current: 2 x 100uA Card decodes 10 address lines, 8 Bit I/O only I/O address # 1: block start = 0x3f8 block length = 0x8 IRQ modes: Level IRQs: 2 3 4 5 7 9 10 11 12 15 Max twin cards = 0 Misc attr: (Audio-BVD2) (Power down supported) Tuple #13, code = 0x1b (Configuration entry), length = 7 000: 25 08 aa 60 f8 02 07 Config index = 0x25 Card decodes 10 address lines, 8 Bit I/O only I/O address # 1: block start = 0x2f8 block length = 0x8 Tuple #14, code = 0x1b (Configuration entry), length = 7 000: 35 08 aa 60 e8 03 07 Config index = 0x35 Card decodes 10 address lines, 8 Bit I/O only I/O address # 1: block start = 0x3e8 block length = 0x8 Tuple #15, code = 0x1b (Configuration entry), length = 7 000: 45 08 aa 60 e8 02 07 Config index = 0x5(default) Card decodes 10 address lines, 8 Bit I/O only I/O address # 1: block start = 0x2e8 block length = 0x8 Tuple #16, code = 0xff (Terminator), length = 0 2 slots found -------------------------------------------------------------------------- PC-card standard says that Multifunction CIS has CISTPL_LONGLINK_MFC (0x06) tupple. According to this output, this card has no 0x06 longlink tupples in it. Morever, it says that this card is "Serial port/modem" only (ID of multifunction card is 0). I think that this card is NOT compatible with PC-card standard multifunction support. Therefore, some sort of dirty hack is needed. According to tupple #11, configuration register of a function (maybe it's serial function, because this card is can be used as serial card without modification of FreeBSD PC-card support) is placed at 0x8020. Looked round attribute memory 0x8020 by pccardc rdattr, I found: -------------------------------------------------------------------------- 7fc0: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 7fd0: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 7fe0: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 7ff0: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 8000: 40 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 8010: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 8020: 45 00 80 00 22 00 00 00 00 00 00 00 00 00 00 00 8030: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 8040: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 8050: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 8060: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 -------------------------------------------------------------------------- 0x8020 seems to be configuration registers of PC-card, but 0x8000 also seems to be configuration registers. So, I think that 0x8000 is configuration registers of another function (maybe Ethernet). Also looking round attribute memory of this CIS, I found another garbage bytes follows normal CIS tupples. -------------------------------------------------------------------------- 0160: 02 00 07 00 1b 00 07 00 35 00 08 00 aa 00 60 00 0170: e8 00 03 00 07 00 1b 00 07 00 45 00 08 00 aa 00 0180: 60 00 e8 00 02 00 07 00 ff 00 ff 00 00 00 00 00 ~~ end of CIS (0xff) 0190: 06 00 af 00 ba 00 47 00 0c 00 57 00 a4 00 58 00 01a0: 00 00 01 00 07 00 57 00 04 00 07 00 90 00 b3 00 01b0: 1f 00 7a 00 af 00 30 00 af 00 00 00 4b 00 8f 00 01c0: ef 00 11 00 c9 00 34 00 00 00 04 00 00 00 30 00 01d0: af 00 57 00 a5 00 ab 00 77 00 f6 00 2f 00 07 00 01e0: 04 00 00 00 00 00 af 00 b5 00 91 00 27 00 f6 00 01f0: 2f 00 04 00 04 00 af 00 c7 00 d6 00 27 00 b6 00 0200: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 0210: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 0220: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 0230: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 0240: ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 ff 00 -------------------------------------------------------------------------- I don't know how to decode this enigma, and perhaps it's really a garbage :-<. If you can decode this, please tell me the truth :-). Anyway, according to PC-card standard, multifunction IRQ shareing can be identified by searching all function's Intr bit of configuration option register. PC-card support of FreeBSD already has pcicintr() first level interrupt handler. So I think it's not difficult to add the code that identify which function's Intr bit, dispatch the interrupt request to the appropriate second-level interrupt handler. Moreover, according to PC-card standard bit 2-0 of config index of multifunction card means, bit 0: enable/disable this function bit 1: number of I/O addresss used bit 2: enables IREQ# routing So, default config index 0x05 means that it enable this function and enable IREQ# routing. All config index has 0x05 bits, so I 'feel' this card partly compatible with PC-card standard. I'll hack this card and try to write first level interrupt handler this month. -- HOSOKAWA, Tatsumi Network Technology Center Keio University hosokawa@ntc.keio.ac.jp