]> git.openfabrics.org - ~shefty/rdma-dev.git/commitdiff
PCI: Update MCP55 quirk to not affect non HyperTransport variants
authorNeil Horman <nhorman@tuxdriver.com>
Wed, 8 Dec 2010 14:47:48 +0000 (09:47 -0500)
committerJesse Barnes <jbarnes@virtuousgeek.org>
Thu, 16 Dec 2010 22:07:31 +0000 (14:07 -0800)
I wrote this quirk awhile ago to properly setup MCP55 chips on hypertransport
busses so that interrupts reached whatever cpu happend to boot the kdump kernel.
while that works well, it was recently shown to me that a a non-hypertransport
variant of the MCP55 exists, and on those system the register that this quirk
manipulates causes hangs if you write to it.  Since the quirk was only meant to
handle errors found on MCP55 chips that have a HT interface, this patch adds a
filter to make sure the chip is an HT capable before making the needed register
adjustment.  This lets the broken MCP55s work with kdump while not breaking the
non-HT variants.

Resolves https://bugzilla.kernel.org/show_bug.cgi?id=23952

Tested successfully by the reporter and myself.

Cc: stable@kernel.org
Reported-by: Mathieu BĂ©rard <mathieu@mberard.eu>
Acked-by: Vivek Goyal <vgoyal@redhat.com>
Signed-off-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org>

index 6f9350cabbd51d1574ec52bcf287d4065f8b3c36..313c0bda0a8dbc94e650a8321d80f8818ad82633 100644 (file)
@@ -2329,6 +2329,9 @@ static void __devinit nvbridge_check_legacy_irq_routing(struct pci_dev *dev)
        u32 cfg;
+       if (!pci_find_capability(dev, PCI_CAP_ID_HT))
+               return;
        pci_read_config_dword(dev, 0x74, &cfg);
        if (cfg & ((1 << 2) | (1 << 15))) {