Date: Mon, 26 Oct 2009 12:31:12 -0700 From: Sean Bruno <sean.bruno@dsl-only.net> To: Fabio <info@maconnect.ch> Cc: freebsd-firewire <freebsd-firewire@freebsd.org> Subject: [Fwd: Re: firewire issue] Message-ID: <1256585472.2607.4.camel@Lappy>
next in thread | raw e-mail | index | archive | help
[-- Attachment #1 --] Look over this patch from Andreas. I haven't done anything with it as I don't have the h/w to test with. Sean [-- Attachment #2 --] Received: (qmail 3972 invoked from network); 25 May 2009 14:12:35 -0700 Received: from mx1.pdx.net (HELO sodium.pdx.net) (69.64.224.81) by iron2.pdx.net with (AES256-SHA encrypted) SMTP; 25 May 2009 14:12:35 -0700 Received: from mail.fgznet.ch ([81.92.96.47] helo=smtp.fgznet.ch) by sodium.pdx.net with esmtp (Exim 4.50) id 1M8hTB-0007xn-Te for sean.bruno@dsl-only.net; Mon, 25 May 2009 14:12:37 -0700 Received: from deuterium.andreas.nets ([91.190.8.131]) by smtp.fgznet.ch (8.13.8/8.13.8/Submit_SMTPAUTH) with ESMTP id n4PLCPio034377 for <sean.bruno@dsl-only.net>; Mon, 25 May 2009 23:12:26 +0200 (CEST) (envelope-from andreast-list@fgznet.ch) Message-ID: <4A1B09B9.6030706@fgznet.ch> Date: Mon, 25 May 2009 23:12:25 +0200 From: Andreas Tobler <andreast-list@fgznet.ch> User-Agent: Thunderbird 2.0.0.17 (Macintosh/20080914) MIME-Version: 1.0 To: Sean Bruno <sean.bruno@dsl-only.net> Subject: Re: firewire issue References: <4A11C3B8.2030705@fgznet.ch> <1242696995.32116.3.camel@localhost.localdomain> <4A123A43.6040200@fgznet.ch> <1243022304.3373.17.camel@localhost.localdomain> <4A18610D.9030201@fgznet.ch> <1243197822.8271.1.camel@localhost.localdomain> <4A19B49F.5040109@fgznet.ch> <1243198858.8271.2.camel@localhost.localdomain> <4A1AF89F.8080407@fgznet.ch> <1243282213.3317.4.camel@localhost.localdomain> <4A1B0002.8060106@fgznet.ch> <1243283875.3317.5.camel@localhost.localdomain> In-Reply-To: <1243283875.3317.5.camel@localhost.localdomain> Content-Type: multipart/mixed; boundary="------------070302000605090708050806" X-Scanned-By: MIMEDefang 2.64 on 81.92.96.47 X-Spam-Checker-Version: SpamAssassin 3.2.0-pdx (2007-05-01) on sodium.pdx.net X-Spam-Status: No, hits=0.0, required=7 tests=BAYES_50=0.001 autolearn=no version=3.2.0 X-Spam-Level: / This is a multi-part message in MIME format. --------------070302000605090708050806 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Sean Bruno wrote: > On Mon, 2009-05-25 at 22:30 +0200, Andreas Tobler wrote: >> Sean Bruno wrote: >>> On Mon, 2009-05-25 at 21:59 +0200, Andreas Tobler wrote: >>>> Sean Bruno wrote: >>>>> On Sun, 2009-05-24 at 22:57 +0200, Andreas Tobler wrote: >>>>>> Sean Bruno wrote: >>>>>>> On Sat, 2009-05-23 at 22:48 +0200, Andreas Tobler wrote: >>>>>>>> Sean Bruno wrote: >>>>>>>>> On Tue, 2009-05-19 at 06:49 +0200, Andreas Tobler wrote: >>>>>>>>>> Sean Bruno wrote: >>>>>>>>>>> Hey ... do you have the IEEE firewire specifications? >>>>>>>>>>> >>>>>>>>>>> I have them if you would like a copy. >>>>>>>>>> Yes, please. >>>>>>>>>> >>>>>>>>>> Andreas >>>>>>>>> http://consultcsg.com/firewire.tar.gz >>>>>>>>> >>>>>>>>> Everything I have. >>>>>>>> Thank you! >>>>>>>> >>>>>>>> The docs confirmed that it is not a protocol issue. >>>>>>>> Here I fight with silicon, I need to find out why the bus reset doesn't >>>>>>>> behave as expected. >>>>>>>> My current mods let me discover the attached fw disk. Unfortunately a >>>>>>>> bus reset kills further operation. >>>>>>>> >>>>>>>> So I have to investigate in this direction... >>>>>>>> >>>>>>>> Anyway, thanks again for the standards. >>>>>>>> >>>>>>>> Regards, >>>>>>>> Andreas >>>>>>> I would *assume* that linux had dealt with this. However, I don't see >>>>>>> any other FreeBSD users complaining about their G3's ... Is this problem >>>>>>> specific to the Mac that you have, or is affecting all of the G3's? >>>>>> I don't know if Linux finally dealt with this issue. I asked the fw guy >>>>>> on the Linux side and he couldn't confirm if the issue is solved. I need >>>>>> to ask BenH, the chief crack. >>>>>> >>>>>> It only affects the machines with >>>>>> #define FW_DEVICE_UNINORTH_V1 (0x0018 << 16) >>>>>> and >>>>>> #define FW_VENDORID_APPLE 0x106b >>>>>> >>>>>> It is an old machine, but this hardware really rocks under *BSD iso. Darwin. >>>>>> >>>>>> And true, it is not lifetime critical :) >>>>>> >>>>>> >>>>>> It is more a prove of concept. >>>>>> >>>>>> Andreas >>>>> Interesting. Ask the freebsd-hackers list if anyone can reproduce the >>>>> issue on the same hardware. >>>> hehe, I doubt, but it might be chance. >>>> >>>> Btw, I noticed a panic when trying my mods with your latest patch you >>>> sent me, the one for sparc64. >>>> My mods are only related to byteswapping of packets. And there is no >>>> device attached. >>>> >>>> I have to type the panic by hand: >>>> >>>> fwohci0: initiate bus reset >>>> fwohci0: fwphy_rddata:: 0x1, retry=6 >>>> Sleeping thread (tid 100069, pid 840) owns a non-sleepable lock >>>> mi_switch >>>> sleepq_switch >>>> sleepq_timedwait >>>> _sleep >>>> pause >>>> fwphy_rddata >>>> fwphy_wrdata >>>> fwohci_ibr >>>> firewire_attach >>>> device_attach >>>> device_probe_and_attach >>>> fwohci_pci_add_child >>>> firewire_identify >>>> .... >>>> >>>> Any blinking leds on your side? >>>> >>>> I wonder why fwohci_ibr is coming in that early? >>>> >>>> Is this correct? >>>> >>>> Andreas >>>> >>> Ok, so your modifications without my changes for the card you have in >>> your sparc64 don't panic? That would be wierd. >> Aehm, I meant to say that your patch and my mods produce this panic on >> the iMac. The sparc64 is in sleep mode until I get the second cpu. And I >> doubt that my mods will panic the sparc64. >> >> I tested my mods against a newer PowerBook and an i386 and I took care >> that I do not break existing functionalitiy. Also, my mods are towards >> current svn. I only wondered if your patch might help here in my imac case. >> >>> Also, can you send me the pciconf -lv for the sparc64 board that I >>> changed the code for? >> I will, but as said, currently it is shutdown. >> >> Andreas > > > Ah. Ok, please send me your current diff and I'll look at it. Attached, I'm not happy with, but it is wip. The mod in the firewire.c where I don't set the max_nod to -1 is necessary to prohibit an infinite loop. The BusReset sets the max_node to -1 and currently I can't set it to a meaningful value. Below the messages w/o debug. the "< ------ this one" bothers me. There is one disk attached. firewire and sbp loaded via kldload. So far it looks correct to me except the bus resets? Andreas May 25 23:02:04 imacb kernel: fwohci0: <Apple UniNorth, v1> mem 0xf5000000-0xf5000fff irq 40 at device 14.0 on pci2 May 25 23:02:04 imacb kernel: fwohci0: [ITHREAD] May 25 23:02:04 imacb kernel: fwohci0: Uninorth v1 May 25 23:02:04 imacb kernel: fwohci0: OHCI version 1.0 (ROM=0) May 25 23:02:04 imacb kernel: fwohci0: No. of Isochronous channels is 4. May 25 23:02:04 imacb kernel: fwohci0: EUI64 00:30:65:ff:fe:aa:7c:d0 May 25 23:02:04 imacb kernel: fwohci0: Phy 1394a available S400, 2 ports. May 25 23:02:04 imacb kernel: fwohci0: Link S400, max_rec 2048 bytes. May 25 23:02:04 imacb kernel: firewire0: <IEEE1394(FireWire) bus> on fwohci0 May 25 23:02:04 imacb kernel: fwohci0: Initiate bus reset May 25 23:02:04 imacb kernel: fwohci0: fwohci_intr_core: BUS reset status: 0xffffffff May 25 23:02:04 imacb kernel: fwohci0: fwohci_intr_core: enable bus reset interrupt May 25 23:02:04 imacb kernel: fwohci0: fwohci_intr_core: node_id=0x00000001, SelfID Count=1, CYCLEMASTER mode May 25 23:02:04 imacb kernel: firewire0: 2 nodes, maxhop <= 1 cable IRM irm(1) (me) May 25 23:02:04 imacb kernel: firewire0: bus manager 1 May 25 23:02:04 imacb kernel: fwohci0: fwohci_intr_core: BUS reset status: 0x6 < ------ this one May 25 23:02:04 imacb kernel: fwohci0: txd err= 3 miss Ack err May 25 23:02:05 imacb kernel: fwohci0: fwohci_intr_core: BUS reset status: 0x5 < ------ this one May 25 23:02:05 imacb kernel: firewire0: fw_explore_node: Pre 1394a-2000 detected May 25 23:02:05 imacb kernel: firewire0: New S400 device ID:0050770e00002fd8 May 25 23:02:13 imacb kernel: sbp0: <SBP-2/SCSI over FireWire> on firewire0 May 25 23:02:13 imacb kernel: sbp0: sbp_show_sdev_info: sbp0:0:0: ordered:1 type:14 EUI:0050770e00002fd8 node:0 speed:2 maxrec:8 May 25 23:02:13 imacb kernel: sbp0: sbp_show_sdev_info: sbp0:0:0 'Prolific PL3507 Combo Device' '(1394 ATAPI,Rev 1.00)' '012804' May 25 23:02:14 imacb kernel: fwohci0: fwohci_intr_core: BUS reset status: 0x7 < ---- this one --------------070302000605090708050806 Content-Type: text/plain; x-mac-type="0"; x-mac-creator="0"; name="fw_20090525.diff" Content-Disposition: inline; filename="fw_20090525.diff" Content-Transfer-Encoding: 7bit Index: fwohci_pci.c =================================================================== --- fwohci_pci.c (revision 192695) +++ fwohci_pci.c (working copy) @@ -196,6 +196,10 @@ device_set_desc(dev, "Apple UniNorth"); return BUS_PROBE_DEFAULT; } + if (id == (FW_VENDORID_APPLE | FW_DEVICE_UNINORTH_V1)) { + device_set_desc(dev, "Apple UniNorth, v1"); + return BUS_PROBE_DEFAULT; + } if (id == (FW_VENDORID_LUCENT | FW_DEVICE_FW322)) { device_set_desc(dev, "Lucent FW322/323"); return BUS_PROBE_DEFAULT; @@ -285,6 +289,10 @@ fwohci_softc_t *sc = device_get_softc(self); int err; int rid; + uint32_t id; + + sc->old_uninorth = 0; + #if defined(__DragonFly__) || __FreeBSD_version < 500000 int intr; /* For the moment, put in a message stating what is wrong */ @@ -383,6 +391,12 @@ return (ENOMEM); } + id = pci_get_devid(self); + if (id == (FW_VENDORID_APPLE | FW_DEVICE_UNINORTH_V1)) { + sc->old_uninorth = 1; + device_printf(self, "Uninorth v1\n"); + } + err = fwohci_init(sc, self); if (err) { Index: firewire.c =================================================================== --- firewire.c (revision 192695) +++ firewire.c (working copy) @@ -77,7 +77,7 @@ struct crom_chunk hw; }; -int firewire_debug=0, try_bmr=1, hold_count=0; +int firewire_debug=2, try_bmr=1, hold_count=0; SYSCTL_INT(_debug, OID_AUTO, firewire_debug, CTLFLAG_RW, &firewire_debug, 0, "FireWire driver debug flag"); SYSCTL_NODE(_hw, OID_AUTO, firewire, CTLFLAG_RD, 0, "FireWire Subsystem"); @@ -632,7 +632,7 @@ CSRARC(fc, TOPO_MAP + 8) = 0; fc->irm = -1; - fc->max_node = -1; + /*fc->max_node = -1; */ for(i = 2; i < 0x100/4 - 2 ; i++){ CSRARC(fc, SPED_MAP + i * 4) = 0; Index: fwohci.c =================================================================== --- fwohci.c (revision 192695) +++ fwohci.c (working copy) @@ -79,6 +79,7 @@ #undef OHCI_DEBUG static int nocyclemaster = 0; +static int old_uninorth = 0; int firewire_phydma_enable = 1; SYSCTL_DECL(_hw_firewire); SYSCTL_INT(_hw_firewire, OID_AUTO, nocyclemaster, CTLFLAG_RW, &nocyclemaster, 0, @@ -266,6 +267,15 @@ d_ioctl_t fwohci_ioctl; +#if BYTE_ORDER == BIG_ENDIAN +#define FWOHCI_DMA_READ_UNI(x) \ + old_uninorth ? (x) : FWOHCI_DMA_READ(x) +#define FWOHCI_DMA_WRITE_UNI(x,y) \ + old_uninorth ? ((x) = (y)) : FWOHCI_DMA_WRITE(x,y) +#else +#define FWOHCI_DMA_READ_UNI(x) FWOHCI_DMA_READ(x) +#define FWOHCI_DMA_WRITE_UNI(x,y) FWOHCI_DMA_WRITE(x,y) +#endif /* * Communication with PHY device */ @@ -624,6 +634,8 @@ return (ENXIO); } + old_uninorth = sc->old_uninorth; + /* Available Isochronous DMA channel probe */ OWRITE(sc, OHCI_IT_MASK, 0xffffffff); OWRITE(sc, OHCI_IR_MASK, 0xffffffff); @@ -952,7 +964,7 @@ if (tcode == FWTCODE_WREQQ || tcode == FWTCODE_RRESQ) hdr_len = 12; for (i = 0; i < hdr_len/4; i ++) - FWOHCI_DMA_WRITE(ld[i], ld[i]); + FWOHCI_DMA_WRITE(ld[i], ld[i]); #endif again: @@ -1811,7 +1823,7 @@ fwohci_dump_intr(struct fwohci_softc *sc, uint32_t stat) { if(stat & OREAD(sc, FWOHCI_INTMASK)) - device_printf(fc->dev, "INTERRUPT < %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s> 0x%08x, 0x%08x\n", + device_printf(sc->fc.dev, "INTERRUPT < %s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s> 0x%08x, 0x%08x\n", stat & OHCI_INT_EN ? "DMA_EN ":"", stat & OHCI_INT_PHY_REG ? "PHY_REG ":"", stat & OHCI_INT_CYC_LONG ? "CYC_LONG ":"", @@ -1845,11 +1857,13 @@ FW_GLOCK_ASSERT(fc); if ((stat & OHCI_INT_PHY_BUS_R) && (fc->status != FWBUSRESET)) { - fc->status = FWBUSRESET; + /* Disable bus reset interrupt until sid recv. */ OWRITE(sc, FWOHCI_INTMASKCLR, OHCI_INT_PHY_BUS_R); - device_printf(fc->dev, "%s: BUS reset\n", __func__); + device_printf(fc->dev, "%s: BUS reset status: 0x%x\n", + __func__, fc->status); + fc->status = FWBUSRESET; OWRITE(sc, FWOHCI_INTMASKCLR, OHCI_INT_CYC_LOST); OWRITE(sc, OHCI_LNKCTLCLR, OHCI_CNTL_CYCSRC); @@ -1863,6 +1877,8 @@ } if (stat & OHCI_INT_PHY_SID) { /* Enable bus reset interrupt */ + device_printf(fc->dev, "%s: enable bus reset interrupt\n", + __func__); OWRITE(sc, FWOHCI_INTSTATCLR, OHCI_INT_PHY_BUS_R); OWRITE(sc, FWOHCI_INTMASK, OHCI_INT_PHY_BUS_R); @@ -2037,7 +2053,7 @@ return; } for (i = 0; i < plen / 4; i ++) - buf[i] = FWOHCI_DMA_READ(sc->sid_buf[i+1]); + buf[i] = FWOHCI_DMA_READ_UNI(sc->sid_buf[i+1]); /* pending all pre-bus_reset packets */ fwohci_txd(sc, &sc->atrq); @@ -2076,9 +2092,12 @@ "device physically ejected?\n"); return (FILTER_STRAY); } - if (stat) - OWRITE(sc, FWOHCI_INTSTATCLR, stat & ~OHCI_INT_PHY_BUS_R); + + if (stat) { + OWRITE(sc, FWOHCI_INTSTATCLR, stat & ~OHCI_INT_PHY_BUS_R); + } + stat &= sc->intmask; if (stat == 0) return (FILTER_STRAY); @@ -2657,7 +2676,7 @@ int i; #endif - ld0 = FWOHCI_DMA_READ(fp->mode.ld[0]); + ld0 = FWOHCI_DMA_READ_UNI(fp->mode.ld[0]); #if 0 printf("ld0: x%08x\n", ld0); #endif @@ -2690,7 +2709,7 @@ } #if BYTE_ORDER == BIG_ENDIAN for(i = 0; i < slen/4; i ++) - fp->mode.ld[i] = FWOHCI_DMA_READ(fp->mode.ld[i]); + fp->mode.ld[i] = FWOHCI_DMA_READ_UNI(fp->mode.ld[i]); #endif return(hlen); } @@ -2884,7 +2903,7 @@ printf("nvec == 0\n"); /* DMA result-code will be written at the tail of packet */ - stat = FWOHCI_DMA_READ(*(uint32_t *)(ld - sizeof(struct fwohci_trailer))); + stat = FWOHCI_DMA_READ_UNI(*(uint32_t *)(ld - sizeof(struct fwohci_trailer))); #if 0 printf("plen: %d, stat %x\n", plen ,stat); Index: fwohcireg.h =================================================================== --- fwohcireg.h (revision 192695) +++ fwohcireg.h (working copy) @@ -73,6 +73,7 @@ #define FW_DEVICE_R5C552 (0x0552 << 16) #define FW_DEVICE_PANGEA (0x0030 << 16) #define FW_DEVICE_UNINORTH (0x0031 << 16) +#define FW_DEVICE_UNINORTH_V1 (0x0018 << 16) #define FW_DEVICE_AIC5800 (0x5800 << 16) #define FW_DEVICE_FW322 (0x5811 << 16) #define FW_DEVICE_7007 (0x7007 << 16) Index: fwohcivar.h =================================================================== --- fwohcivar.h (revision 192695) +++ fwohcivar.h (working copy) @@ -75,6 +75,7 @@ struct task fwohci_task_sid; struct task fwohci_task_dma; int cycle_lost; + int old_uninorth; } fwohci_softc_t; void fwohci_intr (void *arg); --------------070302000605090708050806--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1256585472.2607.4.camel>