From mboxrd@z Thu Jan 1 00:00:00 1970 From: Woody Suwalski Subject: PATCH: cyber2010 framebuffer on ARM Netwinder fix... Date: Tue, 9 Jan 2007 12:26:55 -0500 Message-ID: <45A3D05F.7060409@xandros.com> References: <459D368F.2030204@xandros.com> Mime-Version: 1.0 Content-Type: multipart/mixed; boundary="------------030907010001090908040503" Return-path: In-Reply-To: <20070105152224.GQ15109@deprecation.cyrius.com> References: References: <20070104175352.GC15109@deprecation.cyrius.com> References: References: <20070105152224.GQ15109@deprecation.cyrius.com> Sender: linux-kernel-owner@vger.kernel.org List-Id: , Stuart Anderson , rmk@arm.linux.org.uk, linux-kernel@vger.kernel.org --------------030907010001090908040503 Content-Type: text/plain; charset="ISO-8859-1"; format="flowed" Content-Disposition: inline Martin Michlmayr wrote: > * Stuart Anderson [2007-01-05 09:40]: > >> shark w/o any changes to the kernel. I dug a bit further, both in the >> driver, and in the HW spec for the shark, and discovered that the video >> chip on the shark is connected via the VL bus, not the PCI bus. The >> shark does have a VL-PCI bridge, but there doesn't seem to be anything >> connected to the PCI side of it (which matches what lspci says). The >> function containing the patch in question doesn't appear to even run on >> the shark (there is a VL version that is #ifdef SHARK'd), so I'd have >> to say the patch would have not impact on the shark. >> > > OK, good news. Thanks for checking. Woody, can you submit the patch > (with proper intentation) to linux-fbdev-devel@lists.sourceforge.net > As suggested - I am sending this patch to fbdev-devel.... The Netwinder machines with Cyber2010 crash badly when starting Xserver. The workaround is to disable pci burst option for this revision of video chip. Thanks, Woody --------------030907010001090908040503 Content-Type: text/x-patch; name="netwinder_cyber2010_burst.patch" Content-Disposition: inline; filename="netwinder_cyber2010_burst.patch" The Netwinder machines with Cyber2010 crash badly when starting Xserver. The workaround is to disable pci burst option for this revision of video chip. Signed-off-by: Woody Suwalski --- --- a/drivers/video/cyber2000fb.c 2006-11-29 16:57:37.000000000 -0500 +++ b/drivers/video/cyber2000fb.c 2007-01-09 12:20:01.000000000 -0500 @@ -1539,16 +1539,24 @@ static int cyberpro_pci_enable_mmio(stru /* * Allow the CyberPro to accept PCI burst accesses */ - val = cyber2000_grphr(EXT_BUS_CTL, cfb); - if (!(val & EXT_BUS_CTL_PCIBURST_WRITE)) { - printk(KERN_INFO "%s: enabling PCI bursts\n", cfb->fb.fix.id); - - val |= EXT_BUS_CTL_PCIBURST_WRITE; - - if (cfb->id == ID_CYBERPRO_5000) - val |= EXT_BUS_CTL_PCIBURST_READ; - - cyber2000_grphw(EXT_BUS_CTL, val, cfb); + if (cfb->id == ID_CYBERPRO_2010) + { + printk(KERN_INFO "%s: NOT enabling PCI bursts\n", cfb->fb.fix.id); + } + else + { + val = cyber2000_grphr(EXT_BUS_CTL, cfb); + if (!(val & EXT_BUS_CTL_PCIBURST_WRITE)) { + printk(KERN_INFO "%s: enabling PCI bursts\n", + cfb->fb.fix.id); + + val |= EXT_BUS_CTL_PCIBURST_WRITE; + + if (cfb->id == ID_CYBERPRO_5000) + val |= EXT_BUS_CTL_PCIBURST_READ; + + cyber2000_grphw(EXT_BUS_CTL, val, cfb); + } } return 0; --------------030907010001090908040503--