* protect against invocations of ctype-macros/-functions with negative
character indices by mirroring the corresponding values at negative
array indices (the legal access range of __ctype_b is now [-128..255])
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38708 a95241bf-73f2-0310-859d-f6bbb57e9c96
* avoid trying to overwrite values of a constant structure when
updating the numeric locale data values used by glibc
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38623 a95241bf-73f2-0310-859d-f6bbb57e9c96
the POSIX specs are a bit vague on this (to say the least)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38606 a95241bf-73f2-0310-859d-f6bbb57e9c96
This might help with ACPI shutdown issues, if not this change can be reverted. Not verified as it works on all my machines even without this.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38585 a95241bf-73f2-0310-859d-f6bbb57e9c96
If no Haiku remote disk server is found, try an iSCSI discovery session on
the TFTP server, using the IP address from the boot command or OF options.
All available IPv4 targets are considered, Target Portal Groups are ignored.
RFC 4173 suggests a mechanism that avoids a discovery session by using DHCP;
that requires a compatibly configured DHCP server though and we wouldn't have
access to such data currently anyway. iSCSI is currently used as fallback,
and when it doesn't succeed it falls back to the menu as before.
Resolves ticket #5319.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38537 a95241bf-73f2-0310-859d-f6bbb57e9c96
Add support for both discovery and regular iSCSI sessions. Command and status
sequence numbers do differentiate between session and connection but only
one connection per session is currently supported.
Code is Big Endian for now, so compile it for ppc only.
Based on RFC 3720 ff. Tested against OpenSolaris 2009.06.
Resolves most of ticket #5319.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38536 a95241bf-73f2-0310-859d-f6bbb57e9c96
Don't enqueue a packet whose sequence number matches the first one in queue
or is between the next and the acknowledged on. This would lead to packet
duplication in the queue and leftovers after dequeuing.
Optimize calculation of acknowledgement number for out-of-order reception.
If a single missing packet arrived, this could have resulted in a timeout
depending on the window size.
In case the connection is closed, do dequeue remaining packets. This would
have caused a read error after a unidirectional FIN.
Prepare a dynamic window size calculation.
Do checksum calculation before attempting to trace packet contents. I have
seen a few and couldn't spot an error in the checksum calculation.
More fine-grained control over trace output.
Trace the Maximum Segment Size option if present; break after End of Options.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38535 a95241bf-73f2-0310-859d-f6bbb57e9c96
Pass the desired window size from the socket to the service.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38530 a95241bf-73f2-0310-859d-f6bbb57e9c96
In some error conditions the packet would have been leaked.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38503 a95241bf-73f2-0310-859d-f6bbb57e9c96
Initialize the PRNG seed for the sequence number with real_time_clock_usecs()
rather than system_time(). The remaining uses of system_time() for timeout
calculation match UDP.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38482 a95241bf-73f2-0310-859d-f6bbb57e9c96
Partially revert r38465 and move the code from system_time() into a new
real_time_clock_usecs() function. The system_time() implementation was correct
in relying on of_milliseconds(), as pointed out by Axel. Add the correct
function for the desired functionality.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38481 a95241bf-73f2-0310-859d-f6bbb57e9c96
Fix coding style issues and bump copyright (although there's not much anyway).
No functional changes.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38480 a95241bf-73f2-0310-859d-f6bbb57e9c96
Modelled after UDP, add limited TCP support to the boot net stack. The socket
works by queuing received packets as well as sent packets that have not yet
been ACK'ed. Some known issues are documented, especially there's only limited
congestion control. I.e., we send immediately and in unlimited quantity, thus
its use should be restricted to local networks, and due to a fixed window size
there is potential for our socket being overrun with data packets before they
are read. Some corner cases like wrapping sequence numbers may cause a timeout.
The TCP implementation is based on Andrew S. Tanenbaum's "Computer Networks",
4th ed., as well as lecture notes from Prof. W. Effelsberg, the relevant RFCs
and Wikipedia. The pseudo-random number Galois LFSR used for the sequence
number was suggested by Endre Varga.
Since the code is unlikely to get much smaller, better merge it now so that
subsequent changes get easier to review. No platform actively uses TCP sockets
yet, and the receiving code has been reviewed for endianness issues and should
terminate okay after verifying the checksum if no sockets are open.
Based on a version tested with custom code (#5240) as well as with iSCSI.
Compile-tested boot_loader_openfirmware, pxehaiku-loader with gcc4 and
haiku_loader with gcc2. Closes ticket #5240.
Changes from #5240 proposed patch:
* Various bug fixes related to queuing, some memory leaks fixed.
* Never bump the sequence number when dequeuing a packet. It's done afterwards.
* Don't bump the sequence number again when resending the queue or ACK'ing.
* Aggressively ACK while waiting for packets.
* Don't queue sent ACK-only packets.
* More trace output, esp. for queue inspection.
* Adapted use of TCP header flags to r38434.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38472 a95241bf-73f2-0310-859d-f6bbb57e9c96
system_time() was based on of_milliseconds(), which returns the number of
milliseconds since power-on. This would produce very similar or identical
results for subsequent boots due to limited clock resolution; therefore it
was unsuited as PRNG seed, e.g., for TCP ports.
Try to inquire the RTC device node with get-time to return an improved time
value, if possible.
Closes ticket #6061.
Changes from proposed patch:
* Obtain the RTC handle once and reuse it for each inquiry.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38465 a95241bf-73f2-0310-859d-f6bbb57e9c96
sometimes not.
* This fixes CIDs 1755-1756, and 1505-1506.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38460 a95241bf-73f2-0310-859d-f6bbb57e9c96
make much sense to wake up any waiters either in this case.
* Fixed doubled semaphore deletion.
* Minor cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38418 a95241bf-73f2-0310-859d-f6bbb57e9c96
discovered by Coverity for some reason): commands without arguments would
overwrite memory (in this case the sConsole::arg_count only, though).
* Travis's copyright was missing in blue_screen.cpp.
* Minor cleanup.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38409 a95241bf-73f2-0310-859d-f6bbb57e9c96
Read the EDID from OF and pass it on to the kernel.
Last part of ticket #6105.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38307 a95241bf-73f2-0310-859d-f6bbb57e9c96
Use the OF "screen" device alias to query frame buffer properties. Postpone
the opening of the device as far as possible as it erases the screen output.
Initialize the kernel arguments and hook into generic code to display the
splash screen.
Like on x86, the frame buffer is enabled by default. To disable it, either
press ESC during early boot or, for debugging, make sure in start.cpp that
platform_boot_options() has the BOOT_OPTION_DEBUG_OUTPUT flag set.
Resolves ticket #6105.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38306 a95241bf-73f2-0310-859d-f6bbb57e9c96
Avoid writing to stdout if the frame buffer is enabled.
Part of ticket #6105.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38305 a95241bf-73f2-0310-859d-f6bbb57e9c96
Introduce a non-blocking function for checking keyboard input, and refactor
existing code to share the escape key translations. The comment that key-up and
key-down result in a zero char is confirmed to apply.
If the space bar is pressed, enter the boot menu. If the escape key is pressed,
disable the frame buffer and keep showing the usual debug output. Apparently
the key press must come after console/keyboard initialization but before the
"Welcome to the Haiku bootloader!" line.
Closes ticket #6140.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38304 a95241bf-73f2-0310-859d-f6bbb57e9c96
The frame buffer will need more generic sources than just the text menu.
Adapted from bios_ia32 platform.
Part of ticket #6105.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38301 a95241bf-73f2-0310-859d-f6bbb57e9c96
Debug output calls OpenFirmware methods writing text into the frame buffer.
Depending on the screen mode this leads to garbled display or parts of the
screen overwritten, so mute the output during frame buffer use for now.
A better solution would be to redirect the output elsewhere (e.g., #6168),
so that ConsoleHandle::WriteAt() is not called for the problematic handle
in the first place. That does not work inside the kernel yet.
Part of ticket #6105.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38300 a95241bf-73f2-0310-859d-f6bbb57e9c96
Initialize the frame buffer as not enabled. Add checks for this condition
and for the debug boot option to the video functions.
Code is adapted from bios_ia32 platform. Part of ticket #6105.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38299 a95241bf-73f2-0310-859d-f6bbb57e9c96
This is a leftover from the r25913 build fix. It used to be an out parameter.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38294 a95241bf-73f2-0310-859d-f6bbb57e9c96
This remedies "Current thread pointer [...] is an address we can't read from."
on entering KDL. Closes ticket #6163.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38293 a95241bf-73f2-0310-859d-f6bbb57e9c96
Let setting the memory type to B_MTR_UC succeed.
Add initial support for those memory types in the translation map,
pointed out by Ingo in ticket #5193.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38292 a95241bf-73f2-0310-859d-f6bbb57e9c96
Revert r36886 and fix compilation of insert_virtual_range_to_keep().
The use of void* vs. addr_t matches the surrounding boot loader code
but should probably be revised in favour of addr_t.
create_area() in the kernel wrongly assumed a RAM-backed address range,
which was destined to fail since ranges below the kernel address space
were ignored anyway. Use vm_map_physical_memory() instead.
This fixes a hang once the frame buffer and other resources used by OF
get unmapped. Closes ticket #5193 again.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38291 a95241bf-73f2-0310-859d-f6bbb57e9c96
There are TODO's in here, because the ACPI module needs restructuring and which I will take care of.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38289 a95241bf-73f2-0310-859d-f6bbb57e9c96
which has rather desastrous consequences - thanks to Stippi for the heads
up :-)
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38242 a95241bf-73f2-0310-859d-f6bbb57e9c96
discard mechanism eventually causing the conflicts between the file cache
and the block cache it intended to fix.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38232 a95241bf-73f2-0310-859d-f6bbb57e9c96
this (and other checks) should not be necessary in the first place.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38229 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Also fixed a leak of the same object in team_create_thread_start() I spotted
while looking into it.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38227 a95241bf-73f2-0310-859d-f6bbb57e9c96
* Made an enum out of the mailbox type.
* Rearranged some code to get rid of CID 1328 which was not a bug, though.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38209 a95241bf-73f2-0310-859d-f6bbb57e9c96
(this might also have derefenced a potentially freed vnode).
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@38196 a95241bf-73f2-0310-859d-f6bbb57e9c96
