13 years ago[USB] remove unused material in driver code usb
Balaji Rao [Thu, 21 Aug 2008 17:50:07 +0000 (23:20 +0530)]
[USB] remove unused material in driver code

13 years ago[USB] fix util/
Balaji Rao [Thu, 21 Aug 2008 17:44:26 +0000 (23:14 +0530)]
[USB] fix util/

13 years agoMerge branch 'master' into usb
root [Thu, 21 Aug 2008 17:42:15 +0000 (23:12 +0530)]
Merge branch 'master' into usb

13 years ago[USB] cleanup driver source code
Balaji Rao [Thu, 21 Aug 2008 17:14:01 +0000 (22:44 +0530)]
[USB] cleanup driver source code

13 years ago[USB] OHCI cleanup
Balaji Rao [Thu, 21 Aug 2008 17:10:20 +0000 (22:40 +0530)]
[USB] OHCI cleanup

13 years ago[USB] OHCI cleanup
Balaji Rao [Thu, 21 Aug 2008 17:08:01 +0000 (22:38 +0530)]
[USB] OHCI cleanup

13 years ago[USB] Stop the uhci schedule before cleanup in pci device removal.
Balaji Rao [Thu, 21 Aug 2008 17:05:43 +0000 (22:35 +0530)]
[USB] Stop the uhci schedule before cleanup in pci device removal.

13 years ago[USB] handle ohci_hcd pci device removal cleanly
Balaji Rao [Thu, 21 Aug 2008 17:02:22 +0000 (22:32 +0530)]
[USB] handle ohci_hcd pci device removal cleanly

13 years ago[USB] handle uhci hc pci device removal cleanly.
Balaji Rao [Thu, 21 Aug 2008 16:53:25 +0000 (22:23 +0530)]
[USB] handle uhci hc pci device removal cleanly.

13 years ago[USB] add a usb_hcd_remove_all_devices helper
Balaji Rao [Thu, 21 Aug 2008 16:44:35 +0000 (22:14 +0530)]
[USB] add a usb_hcd_remove_all_devices helper

13 years ago[USB] use proper error handling in set_address and get_configuration
Balaji Rao [Thu, 21 Aug 2008 16:07:24 +0000 (21:37 +0530)]
[USB] use proper error handling in set_address and get_configuration

13 years ago[pxe] Add extra debug messages to the PXE UDP API calls master
Michael Brown [Thu, 21 Aug 2008 15:52:08 +0000 (16:52 +0100)]
[pxe] Add extra debug messages to the PXE UDP API calls

13 years ago[pxe] Fix broadcast transmissions via PXENV_UDP_WRITE
Michael Brown [Thu, 21 Aug 2008 15:50:51 +0000 (16:50 +0100)]
[pxe] Fix broadcast transmissions via PXENV_UDP_WRITE

pxenv_udp_write() needs to specify a net device, otherwise ipv4 is
unable to route broadcast transmissions.

13 years ago[USB] Add a deug statement in pegasus.c
Balaji Rao [Thu, 21 Aug 2008 14:47:03 +0000 (20:17 +0530)]
[USB] Add a deug statement in pegasus.c

13 years ago[USB] Cleanup dm9601
Balaji Rao [Thu, 21 Aug 2008 13:42:27 +0000 (19:12 +0530)]
[USB] Cleanup dm9601

13 years agoMerge branch 'ohci' into usb
root [Thu, 21 Aug 2008 13:28:51 +0000 (18:58 +0530)]
Merge branch 'ohci' into usb

13 years ago[USB] OHCI cleanup
Balaji Rao [Thu, 21 Aug 2008 13:14:40 +0000 (18:44 +0530)]
[USB] OHCI cleanup

13 years ago[USB] OHCI cleanup
Balaji Rao [Thu, 21 Aug 2008 13:13:11 +0000 (18:43 +0530)]
[USB] OHCI cleanup

13 years ago[undi] Fill in ProtType correctly in PXENV_UNDI_ISR
Michael Brown [Wed, 20 Aug 2008 02:21:37 +0000 (03:21 +0100)]
[undi] Fill in ProtType correctly in PXENV_UNDI_ISR

Determine the network-layer packet type and fill it in for UNDI
clients.  This is required by some NBPs such as emBoot's winBoot/i.

This change requires refactoring the link-layer portions of the
gPXE netdevice API, so that it becomes possible to strip the
link-layer header without passing the packet up the network stack.

13 years ago[undi] Work around broken UNDI polling behaviour in winBoot/i
Michael Brown [Wed, 20 Aug 2008 01:02:51 +0000 (02:02 +0100)]
[undi] Work around broken UNDI polling behaviour in winBoot/i

Some dumb NBPs (e.g. emBoot's winBoot/i) never call PXENV_UNDI_ISR
with FuncFlag=PXENV_UNDI_ISR_START; they just sit in a tight polling
loop merrily violating the PXE spec with repeated calls to
PXENV_UNDI_ISR_IN_PROCESS.  Force a extra calls to netdev_poll() to
cope with these out-of-spec clients.

13 years ago[smbios] Print SMBIOS version number in debugging messages
Michael Brown [Tue, 19 Aug 2008 21:01:38 +0000 (22:01 +0100)]
[smbios] Print SMBIOS version number in debugging messages

13 years ago[USB] simplify util/ code to parse USB_ROM macro
Balaji Rao [Tue, 19 Aug 2008 18:40:49 +0000 (00:10 +0530)]
[USB] simplify util/ code to parse USB_ROM macro

13 years ago[USB] Cleanup asix driver
Balaji Rao [Tue, 19 Aug 2008 18:39:32 +0000 (00:09 +0530)]
[USB] Cleanup asix driver

13 years agoMerge branch 'usb' into ohci
root [Tue, 19 Aug 2008 18:28:31 +0000 (23:58 +0530)]
Merge branch 'usb' into ohci

13 years ago[USB] Cleanups in dm9601 driver
Balaji Rao [Tue, 19 Aug 2008 07:54:47 +0000 (13:24 +0530)]
[USB] Cleanups in dm9601 driver

13 years ago[USB] change usbnet.h into dm9601.h as its not used by anyone else.
Balaji Rao [Mon, 18 Aug 2008 12:54:38 +0000 (18:24 +0530)]
[USB] change usbnet.h into dm9601.h as its not used by anyone else.

13 years ago[USB] Remove unused entries in uhci_hcd.h
Balaji Rao [Mon, 18 Aug 2008 07:01:53 +0000 (12:31 +0530)]
[USB] Remove unused entries in uhci_hcd.h

13 years ago[USB] free tds in error path
Balaji Rao [Mon, 18 Aug 2008 06:56:22 +0000 (12:26 +0530)]
[USB] free tds in error path

13 years ago[USB] free tds in error path
Balaji Rao [Mon, 18 Aug 2008 06:54:53 +0000 (12:24 +0530)]
[USB] free tds in error path

13 years ago[pcbios] Support arbitrary splits of the e820 memory map
Michael Brown [Mon, 18 Aug 2008 06:14:27 +0000 (07:14 +0100)]
[pcbios] Support arbitrary splits of the e820 memory map

Allow for an arbitrary number of splits of the system memory map via
INT 15,e820.

Features of the new map-mangling algorithm include:

  Supports random access to e820 map entries.

  Requires only sequential access support from the underlying e820
  map, even if our caller uses random access.

  Empty regions will always be stripped.

  Always terminates with %ebx=0, even if the underlying map terminates
  with CF=1.

  Allows for an arbitrary number of hidden regions, with underlying
  regions split into as many subregions as necessary.

Total size increase to achieve this is 193 bytes.

13 years ago[pcbios] Prepare for multiple splits of hidden e820 memory regions
Michael Brown [Mon, 18 Aug 2008 00:01:45 +0000 (01:01 +0100)]
[pcbios] Prepare for multiple splits of hidden e820 memory regions

Define a list of N allowed memory regions, and split each underlying
e820 region into up to N subregions.  Strip resulting empty regions
out of the map, avoiding using the "return with CF set to strip last
empty region" trick, because it seems that bootmgr.exe in Win2k8 gets
upset if the memory map is terminated with CF set.

This is an intermediate checkin that defines a single allowed memory
region covering the entire 64-bit address space, and uses the existing
map-mangling code on top of the new region-splitting code.  This
sanitises the memory map to the point that Win2k8 is able to boot even
on a system that defines a final zero-length region at the 4GB mark.

I'm checking this in because it may be useful for future debugging
efforts to be able to run with the existing and known-working map
mangling code together with the map sanitisation capabilities of the
new map mangling code.

13 years ago[USB] Fixed various tiny problems identified by mcb30.
Balaji Rao [Sat, 16 Aug 2008 20:33:15 +0000 (02:03 +0530)]
[USB] Fixed various tiny problems identified by mcb30.

13 years ago[USB] Hack to make things work properly.
Balaji Rao [Fri, 15 Aug 2008 09:12:25 +0000 (14:42 +0530)]
[USB] Hack to make things work properly.

13 years ago[util] Allow Option::ROM to understand and modify initialisation entry point
Michael Brown [Fri, 15 Aug 2008 03:10:35 +0000 (04:10 +0100)]
[util] Allow Option::ROM to understand and modify initialisation entry point

Add support for manipulating the jump instruction that forms the
option ROM initialisation entry point, so that can treat
it just like other entry points.

Add support for merging the initialisation entry point (and IBM BOFM
table) to; this is another slightly icky but unfortunately
necessary GPL vs. NDA workaround.  When replaces an entry
point in the original ROM, it now fills in the corresponding entry
point in the merged ROM with the original value; this allows (for
example) a merged initialisation entry point to do some processing and
then jump back to the original entry point.

13 years ago[USB] Kind of works..
Balaji Rao [Thu, 14 Aug 2008 15:58:36 +0000 (21:28 +0530)]
[USB] Kind of works..

13 years ago[settings] Avoid overwriting the start of .text in fetch_string_setting()
Michael Brown [Thu, 14 Aug 2008 02:03:53 +0000 (03:03 +0100)]
[settings] Avoid overwriting the start of .text in fetch_string_setting()

fetch_string_setting() was subtracting one from the length of the
to-be-NUL-terminated buffer in order to obtain the length of the
unterminated buffer to be passed to fetch_setting().  This works
extremely well unless the length of the to-be-NUL-terminated buffer is
zero, at which point we end up giving fetch_setting() a buffer of
length -1UL, thereby inviting it to overwrite as much memory as it

13 years ago[dhcp] Do not restrict minimum retry time for ProxyDHCPREQUEST
Michael Brown [Wed, 13 Aug 2008 20:51:53 +0000 (21:51 +0100)]
[dhcp] Do not restrict minimum retry time for ProxyDHCPREQUEST

The ProxyDHCPREQUEST is a unicast packet, so the first request will
almost always be lost due to not having the IP address in the ARP
cache.  If the minimum retry time is set to one second (as per commit
ff2b6a5), then ProxyDHCP will time out and give up before managing to
successfully transmit a request.

The DHCP timers need to be reworked anyway, so this mild hack is
acceptable for now.

13 years ago[USB] OHCI now works on bare metal. Device probing a bit buggy though.
Balaji Rao [Tue, 12 Aug 2008 18:55:26 +0000 (00:25 +0530)]
[USB] OHCI now works on bare metal. Device probing a bit buggy though.

13 years ago[retry] Added configurable timeouts to retry timer
Andrew Schran [Tue, 12 Aug 2008 00:05:26 +0000 (01:05 +0100)]
[retry] Added configurable timeouts to retry timer

New min_timeout and max_timeout fields in struct retry_timer allow
users of this timer to set their own desired minimum and maximum
timeouts, without being constrained to a single global minimum and
maximum. Users of the timer can still elect to use the default global
values by leaving the min_timeout and max_timeout fields as 0.

13 years ago[iSCSI] Add support for mutual CHAP
Michael Brown [Mon, 11 Aug 2008 02:12:38 +0000 (03:12 +0100)]
[iSCSI] Add support for mutual CHAP

Allow initiator to verify target authentication using CHAP.

13 years ago[libc] Add missing __attribute__ (( format ( printf ) )) to ssnprintf()
Michael Brown [Mon, 11 Aug 2008 01:30:46 +0000 (02:30 +0100)]
[libc] Add missing __attribute__ (( format ( printf ) )) to ssnprintf()

13 years ago[romprefix] Fix ROM image copy on PCI 3.0 BIOSes
Michael Brown [Tue, 5 Aug 2008 22:18:07 +0000 (23:18 +0100)]
[romprefix] Fix ROM image copy on PCI 3.0 BIOSes

Add the missing instructions to the code path that I never tested due
to not having a PCI 3.0 system available.


13 years ago[USB] Bulk also works, kind of.
Balaji Rao [Mon, 4 Aug 2008 22:37:05 +0000 (04:07 +0530)]
[USB] Bulk also works, kind of.

13 years ago[USB] Control transfers now work off OHCI.
Balaji Rao [Mon, 4 Aug 2008 19:55:59 +0000 (01:25 +0530)]
[USB] Control transfers now work off OHCI.

13 years ago[USB] Control transfers work properly as verified by qemu debugging messages.
Balaji Rao [Sun, 3 Aug 2008 19:55:38 +0000 (01:25 +0530)]
[USB] Control transfers work properly as verified by qemu debugging messages.
But its still qemu crashes for some reason.

13 years ago[USB] Introduce ohci_submit_control_urb to handle control urbs.
Balaji Rao [Sat, 2 Aug 2008 11:54:23 +0000 (17:24 +0530)]
[USB] Introduce ohci_submit_control_urb to handle control urbs.

13 years ago[pxe] If no ProxyDHCPACK exists, use DHCPACK for the fake ProxyDHCPACK packet
Michael Brown [Sat, 2 Aug 2008 02:03:42 +0000 (03:03 +0100)]
[pxe] If no ProxyDHCPACK exists, use DHCPACK for the fake ProxyDHCPACK packet

WinPE seems to have a bug that causes it to always use the TFTP server
IP address and filename from the ProxyDHCPACK packet, even if the
ProxyDHCPACK packet doesn't exist.  This causes it to end up
attempting to fetch a file such as


If we don't have a ProxyDHCPACK to use, we pretend that it was a copy
of the DHCPACK packet.  This works around the problem, and hopefully
won't surprise any NBPs.

13 years ago[dhcp] Work around a bug in Altiris RDP
Michael Brown [Fri, 1 Aug 2008 20:49:11 +0000 (21:49 +0100)]
[dhcp] Work around a bug in Altiris RDP

Altiris erroneously cares about the ordering of DHCP options, and will
get confused if we don't construct them in the order it expects.

This is observed (so far) only when attempting to deploy 64-bit Win2k3.

13 years ago[virtio] Add legacy driver for KVM virtio-net
Stefan Hajnoczi [Fri, 1 Aug 2008 15:55:49 +0000 (16:55 +0100)]
[virtio] Add legacy driver for KVM virtio-net

This patch adds support for the virtio-net adapter provided by KVM.

Written by Laurent Vivier <> for Etherboot.
Wrapped as legacy driver for gPXE by Stefan Hajnoczi

13 years agoMerge branch 'usb' into ohci
root [Fri, 1 Aug 2008 12:52:39 +0000 (18:22 +0530)]
Merge branch 'usb' into ohci

13 years ago[USB] Use USB_ROM macro in asix driver.
Balaji Rao [Fri, 1 Aug 2008 12:42:07 +0000 (18:12 +0530)]
[USB] Use USB_ROM macro in asix driver.

13 years ago[USB] Introduce USB_ROM macro to handle auto inclusion of usbnet drivers.
Balaji Rao [Fri, 1 Aug 2008 12:16:38 +0000 (17:46 +0530)]
[USB] Introduce USB_ROM macro to handle auto inclusion of usbnet drivers.

13 years ago[autoboot] Retain initial-slash (if present) when constructing TFTP URIs
Michael Brown [Thu, 31 Jul 2008 15:55:46 +0000 (16:55 +0100)]
[autoboot] Retain initial-slash (if present) when constructing TFTP URIs

When we boot from a DHCP-supplied filename, we previously relied on
the fact that the current working URI is set to tftp://[next-server]/
in order to resolve the filename into a full tftp:// URI.  However,
this process will eliminate the distinction between filenames with and
without initial slashes:

 cwuri="tftp://" filename="vmlinuz"  => URI="tftp://"
 cwuri="tftp://" filename="/vmlinuz" => URI="tftp://"

This distinction is important for some TFTP servers.  We now
explicitly construct a string of the form


so that a filename with an initial slash will result in a URI
containing a double-slash, e.g.


The TFTP code always strips a single initial slash, and so ends up
presenting the correct path to the server.

URIs entered explicitly by users at the command line must include a
double slash if they want an initial slash presented to the TFTP

  "kernel tftp://"  => filename="vmlinuz"
  "kernel tftp://" => filename="/vmlinuz"

13 years ago[cleanup] Remove obsolete README.cvs file
Michael Brown [Thu, 31 Jul 2008 15:24:16 +0000 (16:24 +0100)]
[cleanup] Remove obsolete README.cvs file

13 years ago[USB] Some enqueue_urb work. Actual enqueing of TDs to be done. Code compiles
Balaji Rao [Thu, 31 Jul 2008 14:21:25 +0000 (19:51 +0530)]
[USB] Some enqueue_urb work. Actual enqueing of TDs to be done. Code compiles
cleanly at this point.

13 years ago[util] Add
Michael Brown [Thu, 31 Jul 2008 05:28:46 +0000 (06:28 +0100)]
[util] Add

This utility is required as a workaround for legal restrictions on
including GPL and non-GPL code within the same expansion ROM image.

While this is not encouraged, we are prepared to accept that
concatenation of ROM images and updating of the ROM header data
structures can be classed as "mere aggregation" within the terms of
the GPL.

If in any doubt, assume that you cannot include GPL and non-GPL code
within the same expansion ROM image.  Contact the Etherboot team for
clarification on your specific circumstances.

13 years ago[util] Add Option::ROM library and rewrite to use it.
Michael Brown [Thu, 31 Jul 2008 04:28:11 +0000 (05:28 +0100)]
[util] Add Option::ROM library and rewrite to use it.

The Option::ROM module provides an easy way to read and edit fields
within option ROM headers.

13 years ago[ftp] Terminate processing after receiving an error
Sergey Vlasov [Wed, 30 Jul 2008 19:27:09 +0000 (20:27 +0100)]
[ftp] Terminate processing after receiving an error

When an error reply (not 1xx, 2xx or 3xx) was received, ftp_reply()
invoked ftp_done() to close connections, but did not return, and the
rest of code in this function could try to send commands to the closed
control connection.

Signed-off-by: Sergey Vlasov <>
13 years ago[ftp] Cope with RETR completion prior to all data received
Michael Brown [Wed, 30 Jul 2008 19:22:49 +0000 (20:22 +0100)]
[ftp] Cope with RETR completion prior to all data received

Based on a patch contributed by Sergey Vlasov <> :

  In my testing with "qemu -net user" the 226 response to RETR was
  often received earlier than final packets of the data connection;
  this caused the received file to become truncated without any error
  indication.  Fix this by adding an intermediate state FTP_TRANSFER
  between FTP_RETR and FTP_QUIT, so that the transfer is considered to
  be complete only when both the end of data connection is encountered
  and the final reply to the RETR command is received.

13 years ago[romprefix] Update PCI ROM structure to PCI 3.0
Michael Brown [Wed, 30 Jul 2008 18:57:46 +0000 (19:57 +0100)]
[romprefix] Update PCI ROM structure to PCI 3.0

13 years ago[prefix] Reasonable value for lkrn initrd_addr_max
Stefan Hajnoczi [Wed, 30 Jul 2008 18:45:48 +0000 (19:45 +0100)]
[prefix] Reasonable value for lkrn initrd_addr_max

H. Peter Anvin <> sent word that Sergey Vlasov
<> discovered gPXE lkrn images fail to load in SYSLINUX
3.70 because we have initrd_addr_max zeroed.  This patch sets the same
value as the Linux kernel.

Also change the header jmp instruction to use a hardcoded opcode value
like Linux does.  Just in case the assembler decides to use a three-byte
instruction instead of the desired two-byte jmp.

13 years ago[USB] More OHCI Code. ohci_start.
Balaji Rao [Wed, 30 Jul 2008 12:47:16 +0000 (18:17 +0530)]
[USB] More OHCI Code. ohci_start.

13 years ago[smbios] Add named settings for manufacturer, product name, and serial no.
Michael Brown [Sun, 27 Jul 2008 00:18:55 +0000 (01:18 +0100)]
[smbios] Add named settings for manufacturer, product name, and serial no.

13 years ago[cmdline] Add setting expansion using ${...} syntax
Michael Brown [Sun, 27 Jul 2008 00:08:20 +0000 (01:08 +0100)]
[cmdline] Add setting expansion using ${...} syntax

Allow settings to be expanded in arbitrary commands, such as


Also add the "echo" command, as being the easiest way to test this

13 years ago[USB] Introduce ohci_reset_prot function.
Balaji Rao [Fri, 25 Jul 2008 10:37:10 +0000 (16:07 +0530)]
[USB] Introduce ohci_reset_prot function.

13 years agoMerge branch 'usb' into ohci
root [Fri, 25 Jul 2008 09:55:58 +0000 (15:25 +0530)]
Merge branch 'usb' into ohci

13 years ago[USB] Detect devices on ports.
Balaji Rao [Fri, 25 Jul 2008 09:40:17 +0000 (15:10 +0530)]
[USB] Detect devices on ports.

13 years ago[ui] Add progress dots while waiting on any foreground job
Michael Brown [Thu, 24 Jul 2008 19:08:31 +0000 (20:08 +0100)]
[ui] Add progress dots while waiting on any foreground job

Print one dot per second while waiting in monojob.c (e.g. for DHCP,
for file downloads, etc.), to inform user that the system has not
locked up.

Patch contributed by Andrew Schran <>, minor
modification by me.

13 years ago[cmdline] Minor tidy-ups to shell_banner.c
Michael Brown [Thu, 24 Jul 2008 19:00:43 +0000 (20:00 +0100)]
[cmdline] Minor tidy-ups to shell_banner.c

13 years ago[cmdline] Added configurable shell banner timeout
Andrew Schran [Thu, 24 Jul 2008 18:59:12 +0000 (19:59 +0100)]
[cmdline] Added configurable shell banner timeout

This change allows the time for which shell banners are displayed to
be configured in the config.h file.  The ability to access the shell
can also be effectively disabled by setting this timeout to zero.

13 years ago[tg3] Added support for tg3-5754.
Andrew Schran [Thu, 24 Jul 2008 18:55:53 +0000 (19:55 +0100)]
[tg3] Added support for tg3-5754.

In tg3_chip_reset(), the PCI_EXPRESS change is taken from the Linux
tg3 driver. I am not sure what exactly it does (it is not documented
in the Linux driver), but it is necessary for the NIC to work

13 years ago[makefile] Suppress "No such file or directory" warnings at start of build
Michael Brown [Thu, 24 Jul 2008 18:45:35 +0000 (19:45 +0100)]
[makefile] Suppress "No such file or directory" warnings at start of build

Use "-include" rather than "include" for the generated Makefile
fragments, in order to suppress the long list of warnings that
otherwise appears at the start of a clean build.

Contributed by Edward Waugh <>

13 years ago[phantom] Add new board type
Michael Brown [Thu, 24 Jul 2008 18:40:10 +0000 (19:40 +0100)]
[phantom] Add new board type

13 years ago[USB] More OHCI code. Initialization code now in place.
Balaji Rao [Thu, 24 Jul 2008 18:38:30 +0000 (00:08 +0530)]
[USB] More OHCI code. Initialization code now in place.

13 years ago[USB] Assign proper names to devices. Fix pegasus driver to use it.
Balaji Rao [Wed, 23 Jul 2008 19:40:12 +0000 (01:10 +0530)]
[USB] Assign proper names to devices. Fix pegasus driver to use it.

13 years ago[USB] Improve TD token fix up logic to handle multiple queued URBs.
Balaji Rao [Wed, 23 Jul 2008 19:22:13 +0000 (00:52 +0530)]
[USB] Improve TD token fix up logic to handle multiple queued URBs.

13 years ago[USB] Move usb device initialization logic from uhci driver into usbcore.
Balaji Rao [Tue, 22 Jul 2008 07:52:56 +0000 (13:22 +0530)]
[USB] Move usb device initialization logic from uhci driver into usbcore.

13 years ago[USB] Reduce UHCI PR-deassert PE-assert interval to 10 us.
Balaji Rao [Tue, 22 Jul 2008 07:33:02 +0000 (13:03 +0530)]
[USB] Reduce UHCI PR-deassert PE-assert interval to 10 us.

13 years ago[USB] Add a 'reset_port' method to the hc_driver struct.
Balaji Rao [Tue, 22 Jul 2008 07:30:43 +0000 (13:00 +0530)]
[USB] Add a 'reset_port' method to the hc_driver struct.

13 years ago[iSCSI] Support Windows Server 2008 direct iSCSI installation
Michael Brown [Thu, 17 Jul 2008 16:45:17 +0000 (17:45 +0100)]
[iSCSI] Support Windows Server 2008 direct iSCSI installation

Add yet another ugly hack to iscsiboot.c, this time to allow the user to
inhibit the shutdown/removal of the iSCSI INT13 device (and the network
devices, since they are required for the iSCSI device to function).

On the plus side, the fact that shutdown() now takes flags to
differentiate between shutdown-for-exit and shutdown-for-boot means that
another ugly hack (to allow returning via the PXE stack on BIOSes that
have broken INT 18 calls) will be easier.

I feel dirty.

13 years ago[phantom] Guard against partially-written status descriptors
Michael Brown [Thu, 17 Jul 2008 14:45:40 +0000 (15:45 +0100)]
[phantom] Guard against partially-written status descriptors

Conjecture: The hardware issues 64-bit DMA writes of status descriptors,
which some PCI bridges seem to split into two 32-bit writes in reverse
order (i.e. dword 1 first).  This means that we sometimes observe a
partial status descriptor.  Add an explicit check to ensure that the
descriptor is complete before processing it.

Also ensure that the RDS consumer counter is incremented only when we
know that we have actually consumed an RX descriptor.

13 years ago[USB] Add a simple debugging statement.
Balaji Rao [Wed, 16 Jul 2008 20:27:36 +0000 (01:57 +0530)]
[USB] Add a simple debugging statement.

13 years ago[USB] Add appropriate timing between PR and PE.
Balaji Rao [Wed, 16 Jul 2008 19:52:24 +0000 (01:22 +0530)]
[USB] Add appropriate timing between PR and PE.

13 years ago[int13] Pairwise swap drive numbers, instead of shifting all drive numbers
Michael Brown [Tue, 15 Jul 2008 22:19:59 +0000 (23:19 +0100)]
[int13] Pairwise swap drive numbers, instead of shifting all drive numbers

Shifting all INT13 drive numbers causes problems on systems that use a
sparse drive number space (e.g. qemu BIOS, which uses 0xe0 for the CD-ROM

The strategy now is:

  Each drive is assigned a "natural" drive number, being the next
  available drive number in the system (based on the BIOS drive count).

  Each drive is accessed using its specified drive number.  If the
  specified drive number is -1, the natural drive number will be used.

  Accesses to the specified drive number will be delivered to the
  emulated drive, masking out any preexisting drive using this number.

  Accesses to the natural drive number, if different, will be remapped to
  the masked-out drive.

The overall upshot is that, for examples:

  System has no drives.  Emulated INT13 drive gets natural number 0x80
  and specified number 0x80.  Accesses to drive 0x80 go to the emulated
  drive, and there is no remapping.

  System has one drive.  Emulated INT13 drive gets natural number 0x81
  and specified number 0x80.  Accesses to drive 0x80 go to the emulated
  drive.  Accesses to drive 0x81 get remapped to the original drive 0x80.

13 years ago[USB] More OHCI code
Balaji Rao [Tue, 15 Jul 2008 19:42:28 +0000 (01:12 +0530)]
[USB] More OHCI code

13 years agoSOme thing
Balaji Rao [Fri, 11 Jul 2008 12:44:22 +0000 (18:14 +0530)]
SOme thing

13 years agoMerge branch 'master' into usb
Balaji Rao [Fri, 11 Jul 2008 01:54:56 +0000 (07:24 +0530)]
Merge branch 'master' into usb



13 years ago[USB] Use SPD (Sort Packet Detect) for control transfers as well. USB 2.0 devices
Balaji Rao [Thu, 10 Jul 2008 03:07:30 +0000 (08:37 +0530)]
[USB] Use SPD (Sort Packet Detect) for control transfers as well. USB 2.0 devices
are now configured.

13 years ago[udp] Avoid reusing auto-allocated ports after connection close.
Michael Brown [Thu, 10 Jul 2008 03:21:10 +0000 (04:21 +0100)]
[udp] Avoid reusing auto-allocated ports after connection close.

This duplicates a fix made to tcp.c in commit

13 years ago[dhcp] Verify DHCP server source port
Michael Brown [Thu, 10 Jul 2008 03:11:09 +0000 (04:11 +0100)]
[dhcp] Verify DHCP server source port

Verifying server ID and DHCP transaction ID is insufficient to
differentiate between DHCPACK and ProxyDHCPACK when the DHCP server and
Proxy DHCP server are the same machine.

13 years ago[USB] Fix uhci port reset code.
Balaji Rao [Thu, 10 Jul 2008 02:57:12 +0000 (08:27 +0530)]
[USB] Fix uhci port reset code.

13 years agoRevert "[USB] Fixed USB 2.0 device detection bug. Moved device initialization and"
Balaji Rao [Thu, 10 Jul 2008 02:52:53 +0000 (08:22 +0530)]
Revert "[USB] Fixed USB 2.0 device detection bug. Moved device initialization and"

This reverts commit b0394265d3a7b9c507bc12a5f697ceca8c3a574a.

13 years ago[forcedeth] Add support for PCI ID 10de:0373
Andrew Schran [Thu, 10 Jul 2008 00:19:08 +0000 (01:19 +0100)]
[forcedeth] Add support for PCI ID 10de:0373

13 years ago[USB] Fixed USB 2.0 device detection bug. Moved device initialization and
Balaji Rao [Wed, 9 Jul 2008 00:58:41 +0000 (06:28 +0530)]
[USB] Fixed USB 2.0 device detection bug. Moved device initialization and
configuration code away from uhci_hcd.c into usbcore.c.

13 years ago[hermon] Add support for MT26418 device
Michael Brown [Tue, 8 Jul 2008 23:53:03 +0000 (00:53 +0100)]
[hermon] Add support for MT26418 device

PCI ID supplied (and tested) by Frederic
Temporelli <>

13 years ago[phantom] Remove warning relating to now-fixed firmware bug
Michael Brown [Tue, 8 Jul 2008 21:37:36 +0000 (22:37 +0100)]
[phantom] Remove warning relating to now-fixed firmware bug

13 years ago[phantom] Add NetXen dual copyright
Michael Brown [Tue, 8 Jul 2008 21:36:07 +0000 (22:36 +0100)]
[phantom] Add NetXen dual copyright

13 years ago[phantom] Initialise RCVPEG before reading MAC addresses
Michael Brown [Tue, 8 Jul 2008 19:59:33 +0000 (20:59 +0100)]
[phantom] Initialise RCVPEG before reading MAC addresses

13 years ago[image] Fail "imgexec"/"boot" if the image to execute is ambiguous
Michael Brown [Tue, 8 Jul 2008 02:50:44 +0000 (03:50 +0100)]
[image] Fail "imgexec"/"boot" if the image to execute is ambiguous

If there is more than one loaded image, refuse to automatically select
the image to execute.  There are at least two possible cases, with
different "correct" answers:

1. User loads image A by mistake, then loads image B and types "boot".
   User wants to execute image B.

2. User loads image A, then loads image B (which patches image A), then
   types "boot".  User wants to execute image A.

If a user actually wants to load multiple images, they must explicitly
specify which image is to be executed.

13 years ago[image] Revert "clear LOADED flag" patch
Michael Brown [Tue, 8 Jul 2008 02:15:02 +0000 (03:15 +0100)]
[image] Revert "clear LOADED flag" patch

Clearing the LOADED flag actually prevents users from doing clever things
such as loading an image, then loading a patch image, then executing the
first image.  (image_exec() checks for IMAGE_LOADED, so this sequence of
operations will fail if the LOADED flag gets cleared.)

This reverts commit 14c080020fb37cb34fe74213393f47bd2ad8b9bc.