Michael Brown [Mon, 19 Nov 2007 20:14:49 +0000 (20:14 +0000)]
Merge branch 'master' of rom.etherboot.org:/gpxe
Michael Brown [Mon, 19 Nov 2007 20:14:20 +0000 (20:14 +0000)]
Add bin/pci_VVVV_DDDD target to allow building ROM images without
needing to know the gPXE internal ROM name.
Marty Connor [Sun, 18 Nov 2007 08:11:02 +0000 (03:11 -0500)]
Remove unnecessary variable.
Michael Brown [Mon, 5 Nov 2007 13:29:05 +0000 (13:29 +0000)]
Work around a bug in the OpenSolaris iSCSI target.
We didn't specify values for MaxRecvDataSegmentLength and
MaxBurstLength (to save space, since we were happy with the
RFC-defined default values of 8kB and 256kB respectively). However,
the OpenSolaris target (incorrectly) assumes default values of zero
for these parameters.
The upshot was that the OpenSolaris target would get stuck in an
endless loop trying to send us the first 512-byte sector, zero bytes
at a time, and would eventually run out of memory and core-dump.
Fixed by explicitly specifying the default values for these two
parameters.
Michael Brown [Sun, 4 Nov 2007 04:27:38 +0000 (04:27 +0000)]
Allow space for the kernel's real-mode .bss. Previously we weren't
allowing any space for this, which makes it surprising that bzImage
loading ever worked.
Michael Brown [Mon, 29 Oct 2007 21:22:00 +0000 (21:22 +0000)]
Add missing attribute(packed) to two structures.
Michael Brown [Mon, 29 Oct 2007 17:22:27 +0000 (17:22 +0000)]
Merge branch 'master' of rom.etherboot.org:/gpxe
Michael Brown [Mon, 29 Oct 2007 17:21:58 +0000 (17:21 +0000)]
Merge branch '3leaf'
Michael Brown [Mon, 22 Oct 2007 15:24:09 +0000 (16:24 +0100)]
Added one more PCI ID
Michael Brown [Wed, 10 Oct 2007 23:43:21 +0000 (00:43 +0100)]
Change ROM names to lower case.
Michael Brown [Wed, 10 Oct 2007 19:12:25 +0000 (20:12 +0100)]
Ensure that empty e820 regions are skipped even at the end of the
memory map. (We achieve this by setting CF on the last entry if it is
zero-length; this avoids the need to look ahead to see at each entry
if the *next* entry would be both the last entry and zero-length).
This fixes the "0kB base memory" error message upon starting Windows
2003 on a SunFire X2100.
Michael Brown [Wed, 10 Oct 2007 19:12:25 +0000 (20:12 +0100)]
Ensure that empty e820 regions are skipped even at the end of the
memory map. (We achieve this by setting CF on the last entry if it is
zero-length; this avoids the need to look ahead to see at each entry
if the *next* entry would be both the last entry and zero-length).
This fixes the "0kB base memory" error message upon starting Windows
2003 on a SunFire X2100.
Michael Brown [Wed, 10 Oct 2007 01:07:18 +0000 (02:07 +0100)]
Add a missing PCI ID that is tested as working.
Michael Brown [Thu, 4 Oct 2007 13:46:56 +0000 (14:46 +0100)]
Force a netdevice poll in net_tx() before attempting to transmit
packet, to work around the problem whereby small TX rings get backed
up because we haven't yet poll()ed for TX completions.
Michael Brown [Thu, 4 Oct 2007 13:46:56 +0000 (14:46 +0100)]
Force a netdevice poll in net_tx() before attempting to transmit
packet, to work around the problem whereby small TX rings get backed
up because we haven't yet poll()ed for TX completions.
Michael Brown [Fri, 28 Sep 2007 00:23:06 +0000 (01:23 +0100)]
Redefine bzimage_exec_context::mem_limit to be the highest permissible
byte, rather than the number of permissible bytes (i.e. subtract one
from the value under the previous definition to get the value under
the new definition).
This avoids integer overflow on 64-bit kernels, where
bzhdr.initrd_addr_max may be 0xffffffffffffffff; under the old
behaviour we set mem_limit equal to initrd_addr_max+1, which meant it
ended up as zero. Kernel loads would fail with ENOBUFS.
Michael Brown [Wed, 26 Sep 2007 16:58:51 +0000 (17:58 +0100)]
Quick and very dirty hack to get r8169 driver working again, prior to
any rewrite.
Michael Brown [Wed, 26 Sep 2007 16:38:36 +0000 (17:38 +0100)]
Standardise DBG_PRINTF()->DBG(), and remove custom definitions for
DBG_PRINTF() and assert().
Michael Brown [Tue, 25 Sep 2007 19:16:32 +0000 (20:16 +0100)]
Switch rm_ss and rm_sp back to being words; it'll make it less
confusing to read the code.
Michael Brown [Tue, 25 Sep 2007 19:14:20 +0000 (20:14 +0100)]
Don't use the "rep ss movsb" trick to copy the RM stack to the PM
stack; it breaks vmxassist.
Michael Brown [Tue, 25 Sep 2007 18:20:38 +0000 (19:20 +0100)]
rm_ss, rm_sp, and pm_esp don't need to be accessed from anywhere
outside of librm.S.
Michael Brown [Tue, 25 Sep 2007 18:06:38 +0000 (19:06 +0100)]
Minor refactoring to eliminate duplication.
Michael Brown [Tue, 25 Sep 2007 17:01:15 +0000 (18:01 +0100)]
Use full protected mode for access to high memory within prefix, to
work around limitations in real-mode virtualisation support on Intel
VT-capable chips.
Michael Brown [Fri, 21 Sep 2007 17:27:51 +0000 (18:27 +0100)]
Added more verbose memory-map debugging.
Added redundant call to get_memmap() in int13_boot() immediately
before jumping to boot sector, to assist in debugging.
Michael Brown [Fri, 21 Sep 2007 17:27:51 +0000 (18:27 +0100)]
Added more verbose memory-map debugging.
Added redundant call to get_memmap() in int13_boot() immediately
before jumping to boot sector, to assist in debugging.
Michael Brown [Fri, 21 Sep 2007 00:27:03 +0000 (01:27 +0100)]
Merge branch 'master' into 3leaf-delivery
Michael Brown [Fri, 21 Sep 2007 00:26:06 +0000 (01:26 +0100)]
Merge branch 'master' of rom.etherboot.org:/gpxe
Michael Brown [Fri, 21 Sep 2007 00:23:57 +0000 (01:23 +0100)]
Infinband requires longer link-layer addresses and headers than Ethernet.
Michael Brown [Fri, 21 Sep 2007 00:19:04 +0000 (01:19 +0100)]
Added defines for Infiniband files (which can't be committed to the
main tree yet, but reserving the error numbers now avoids problems later).
Michael Brown [Mon, 17 Sep 2007 17:57:32 +0000 (18:57 +0100)]
Minor debug message improvement.
Michael Brown [Thu, 13 Sep 2007 00:53:04 +0000 (01:53 +0100)]
Use RFC4390 whenever hardware address exceeds 16 bytes; this allows us
to construct DHCP packets suitable for Infiniband.
Michael Brown [Sat, 15 Sep 2007 17:44:35 +0000 (18:44 +0100)]
Add const attribute to byte-swapping functions
Michael Brown [Sun, 16 Sep 2007 16:39:29 +0000 (17:39 +0100)]
Don't get stuck in an infinite loop on negative integers!
Michael Brown [Sun, 16 Sep 2007 16:36:40 +0000 (17:36 +0100)]
Add fls() for non-constant values.
Michael Brown [Mon, 17 Sep 2007 17:38:04 +0000 (18:38 +0100)]
Moved iobuf.h assertions outside the static inline functions, so that
the assert message's file and line number gives some clue as to the
real location of the problem.
Added similar assertions to list.h.
Michael Brown [Tue, 18 Sep 2007 06:07:04 +0000 (07:07 +0100)]
Reverted driver/net/mlx_ipoib to clean master state
Michael Brown [Tue, 18 Sep 2007 06:03:57 +0000 (07:03 +0100)]
Merge branch 'master' into 3leaf-rewrite
Michael Brown [Tue, 18 Sep 2007 06:02:41 +0000 (07:02 +0100)]
IB can't afford many RX buffers, because the MTU is so large. Reduce
the TCP window to compensate.
Michael Brown [Tue, 18 Sep 2007 05:37:22 +0000 (06:37 +0100)]
Added arbel.c
Michael Brown [Tue, 18 Sep 2007 05:26:31 +0000 (06:26 +0100)]
Remove the last remaining visible hack.
Michael Brown [Tue, 18 Sep 2007 05:06:55 +0000 (06:06 +0100)]
Separated out to a clean new drivers/infiniband directory.
Michael Brown [Tue, 18 Sep 2007 04:46:49 +0000 (05:46 +0100)]
First version that works with SELF_INIT!
Michael Brown [Tue, 18 Sep 2007 04:39:34 +0000 (05:39 +0100)]
Gets most of the way through initialisation.
Michael Brown [Tue, 18 Sep 2007 03:45:55 +0000 (04:45 +0100)]
Added sw2hw_mpt
Michael Brown [Tue, 18 Sep 2007 03:20:41 +0000 (04:20 +0100)]
INIT_HCA block now matches.
Michael Brown [Tue, 18 Sep 2007 02:58:59 +0000 (03:58 +0100)]
Seems to get through the ICM mapping process
Michael Brown [Tue, 18 Sep 2007 01:57:49 +0000 (02:57 +0100)]
ICM allocation code in place; about to start test-and-compare runs.
Michael Brown [Tue, 18 Sep 2007 01:05:25 +0000 (02:05 +0100)]
ICM allocation code *should* now be complete.
Michael Brown [Tue, 18 Sep 2007 00:49:39 +0000 (01:49 +0100)]
Started with the ICM partitioning code.
Michael Brown [Mon, 17 Sep 2007 21:25:46 +0000 (22:25 +0100)]
Restructured bits of initialisation (not testable at present)
Michael Brown [Mon, 17 Sep 2007 20:59:41 +0000 (21:59 +0100)]
Starting the firmware directly now works.
Michael Brown [Mon, 17 Sep 2007 20:14:09 +0000 (21:14 +0100)]
Start migrating the remaining initialisation steps.
Michael Brown [Mon, 17 Sep 2007 18:47:19 +0000 (19:47 +0100)]
Don't use the mailboxless version of the HW2SW_CQ command; it seems to
crash the machine.
Michael Brown [Mon, 17 Sep 2007 18:17:55 +0000 (19:17 +0100)]
Removed more hacks, and adjusted number of queue entries for likely usage.
Michael Brown [Mon, 17 Sep 2007 18:12:06 +0000 (19:12 +0100)]
Remove hacks, and fix leaving the multicast GID.
Michael Brown [Mon, 17 Sep 2007 18:01:10 +0000 (19:01 +0100)]
Added an almost obscene amount of debugging and assertion code while
tracking down a bug that turned out to be a free_iob() used where I
needed a netdev_tx_complete(). This left the freed I/O buffer on the
net device's TX list, with bad, bad consequences later.
Also fixed the bug in question.
Michael Brown [Mon, 17 Sep 2007 17:57:32 +0000 (18:57 +0100)]
Minor debug message improvement.
Michael Brown [Mon, 17 Sep 2007 17:38:04 +0000 (18:38 +0100)]
Moved iobuf.h assertions outside the static inline functions, so that
the assert message's file and line number gives some clue as to the
real location of the problem.
Added similar assertions to list.h.
Michael Brown [Mon, 17 Sep 2007 12:31:13 +0000 (13:31 +0100)]
Some interesting packet corruption happening now.
Michael Brown [Mon, 17 Sep 2007 10:50:25 +0000 (11:50 +0100)]
Broadcast GID is now calculated by IPoIB layer.
Michael Brown [Mon, 17 Sep 2007 09:56:21 +0000 (10:56 +0100)]
Cleaned up some debug messages.
Michael Brown [Mon, 17 Sep 2007 09:39:30 +0000 (10:39 +0100)]
Unicasts seem to be working. :)
Michael Brown [Mon, 17 Sep 2007 07:41:06 +0000 (08:41 +0100)]
Dead code disabling
Michael Brown [Mon, 17 Sep 2007 07:36:56 +0000 (08:36 +0100)]
Obtains a response to the get path record!
Michael Brown [Mon, 17 Sep 2007 05:35:21 +0000 (06:35 +0100)]
Use pkey table access to determine broadcast GID directly.
Michael Brown [Mon, 17 Sep 2007 05:12:33 +0000 (06:12 +0100)]
Prepare for adding a metadata queue to IPoIB
Michael Brown [Mon, 17 Sep 2007 04:12:47 +0000 (05:12 +0100)]
Dead code removal
Michael Brown [Mon, 17 Sep 2007 04:04:58 +0000 (05:04 +0100)]
IPoIB code separated out to ipoib.c.
Michael Brown [Mon, 17 Sep 2007 01:54:15 +0000 (02:54 +0100)]
Read port GID directly using MAD IFC.
Michael Brown [Sun, 16 Sep 2007 23:24:44 +0000 (00:24 +0100)]
Multicast join now works.
Michael Brown [Sun, 16 Sep 2007 21:53:57 +0000 (22:53 +0100)]
Now transmits packets on our own allocated IPoIB queue pair. :)
Michael Brown [Sun, 16 Sep 2007 21:47:45 +0000 (22:47 +0100)]
Almost working with own-queue allocation.
Michael Brown [Sun, 16 Sep 2007 19:54:21 +0000 (20:54 +0100)]
create_qp() and destroy_qp() now written (but not tested).
Michael Brown [Sun, 16 Sep 2007 18:03:24 +0000 (19:03 +0100)]
Started implementing create_qp() and destroy_qp().
Michael Brown [Sun, 16 Sep 2007 16:39:29 +0000 (17:39 +0100)]
Don't get stuck in an infinite loop on negative integers!
Michael Brown [Sun, 16 Sep 2007 16:36:40 +0000 (17:36 +0100)]
Add fls() for non-constant values.
Michael Brown [Sun, 16 Sep 2007 16:25:15 +0000 (17:25 +0100)]
Revert to dev_priv/owner_priv scheme, rather than container_of; it
makes it easier to put the generic allocation code into infiniband.c
Michael Brown [Sun, 16 Sep 2007 13:39:51 +0000 (14:39 +0100)]
destroy_cq() now implemented (not tested).
Michael Brown [Sun, 16 Sep 2007 13:18:17 +0000 (14:18 +0100)]
create_cq() implemented (but not tested).
Michael Brown [Sun, 16 Sep 2007 12:10:15 +0000 (13:10 +0100)]
Match doorbell layout to expected usage.
Michael Brown [Sun, 16 Sep 2007 11:52:58 +0000 (12:52 +0100)]
Started to add code for CQ creation
Michael Brown [Sun, 16 Sep 2007 03:02:20 +0000 (04:02 +0100)]
Rearrange data structures to maximise embedding (and hence minimise
the number of separate allocations that need to be done).
Michael Brown [Sun, 16 Sep 2007 02:13:25 +0000 (03:13 +0100)]
Command interface now reasonably friendly.
Michael Brown [Sun, 16 Sep 2007 00:44:57 +0000 (01:44 +0100)]
First (working) draft of command interface.
Michael Brown [Sat, 15 Sep 2007 22:36:41 +0000 (23:36 +0100)]
Kill off more dead code.
Michael Brown [Sat, 15 Sep 2007 22:33:25 +0000 (23:33 +0100)]
post_recv() now works, and we can pass data on the IPoIB queue pair
using entirely our own code.
Michael Brown [Sat, 15 Sep 2007 20:26:14 +0000 (21:26 +0100)]
Kill off some dead code
Michael Brown [Sat, 15 Sep 2007 20:23:34 +0000 (21:23 +0100)]
Hack up IB structures at start-of-day, instead of on each use.
Michael Brown [Sat, 15 Sep 2007 19:58:29 +0000 (20:58 +0100)]
Direct polling of TX completion queue now works.
Michael Brown [Sat, 15 Sep 2007 17:44:35 +0000 (18:44 +0100)]
Add const attribute to byte-swapping functions
Michael Brown [Sat, 15 Sep 2007 17:44:09 +0000 (18:44 +0100)]
Updated MLX_* accessor macros to use implicit type information.
Michael Brown [Sat, 15 Sep 2007 14:40:35 +0000 (15:40 +0100)]
Started added poll_cq() verb.
Started reworking MLX_EXTRACT(), MLX_POPULATE() etc. to automatically
determine type information.
Michael Brown [Sat, 15 Sep 2007 02:22:04 +0000 (03:22 +0100)]
Map the whole of physical memory
Michael Brown [Sat, 15 Sep 2007 01:53:05 +0000 (02:53 +0100)]
arbel_post_send() has been observed to transmit a packet!
Michael Brown [Sat, 15 Sep 2007 00:35:07 +0000 (01:35 +0100)]
Gets a response out of the hardware. (An error completion, to be precise.)
Michael Brown [Fri, 14 Sep 2007 23:27:09 +0000 (00:27 +0100)]
Prefix arbel-specific functions etc. with arbel_
Michael Brown [Fri, 14 Sep 2007 23:19:38 +0000 (00:19 +0100)]
Now at least compiles
Michael Brown [Fri, 14 Sep 2007 19:29:44 +0000 (20:29 +0100)]
Starting to introduce an Infiniband device abstraction
Michael Brown [Fri, 14 Sep 2007 10:23:06 +0000 (11:23 +0100)]
Remove some dead code
Michael Brown [Fri, 14 Sep 2007 10:10:25 +0000 (11:10 +0100)]
Now handling TX completions in our poll loop.