From 35ab235ff164c138da8f90d0cc1a606fcc7c30d9 Mon Sep 17 00:00:00 2001 From: stansmith Date: Sat, 2 Aug 2008 01:09:11 +0000 Subject: [PATCH] [IBBUS] added readme.txt file to explain where testing is and what needs to be done. git-svn-id: svn://openib.tc.cornell.edu/gen1@1455 ad392aa1-c5ef-ae45-8dd8-e69d62a5ef86 --- branches/winverbs/core/bus/kernel/README.txt | 23 ++++++++++++++++++++ 1 file changed, 23 insertions(+) create mode 100644 branches/winverbs/core/bus/kernel/README.txt diff --git a/branches/winverbs/core/bus/kernel/README.txt b/branches/winverbs/core/bus/kernel/README.txt new file mode 100644 index 00000000..2e9da031 --- /dev/null +++ b/branches/winverbs/core/bus/kernel/README.txt @@ -0,0 +1,23 @@ +[8-1-08] + +ibbus as an upper filter driver for mthca (soon mlx4). +Tested to be functional and working for mthca drivers using etc\mthca.inf on +Server 2003 (multiple mthcas) along with a single mthca for Server 2008 & Vista. +IPoIB has been the test vehicle as I have no IOU accessible; need some help here. + +Since ibbus is a filter driver, the main change is inventing a way for the HCA +of interest to be located when add/removing ports or IOUs. +The concept of a vector of Bus Filter instances, once BFI per HCA, is used to +bind an HCA to a BFI vector slot. The BFI contains pointers to the port_mgr and iou_mgr +for the specific HCA. + +HCA binding is delayed until the 1st port or IOU PNP ADD event occurs. Why? in bus_device_add() +where a BFI is allocated, although not bound to an HCA, the HCA guid was not immediately +available as the HCA driver had not yet used the bus CI interface to register +the HCA guid with the bus driver. I could use some alternative ideas here. + +The current folder is a direct replacement for trunk\core\bus\kernel using the head +of the svn tree. There is still much debug code which will be removed once IOU testing +is completed. + +stan. \ No newline at end of file -- 2.46.0