- due to a merge error, 2 line were missing making all the kernel data area
cache inhibited.
- due to a misunderstanding of "kstsize" units, all but the first page of
the kernel segment table was copyback cached on the '040/'060 which
should have caused sporadic user process segmentation faults or
kernel endless loops on the '060, under heavy load (when lots of
userland pagetables are in-core), although the problem was not yet
observed.
an unrelated bug report. This will make kernel startups a bit more readable
in the presence of unsupported hardware.
Information contributed by Andreas Bussjaeger.
- adjust txhiwat and mindma params a bit
- fixed a couple of incorrectly labeled panic calls
- the "location" was being calculated incorrectly in some cases (forgot
to subtract off MID_RAMBASE). this only caused problem when trying
to change the size of the tx/rx buffers (e.g. to 64KB).
- fixed possible non-aligned DMA burst in the starting byte burst case.
(e.g. if we could DMA 3 bytes, but only have 2 it is not legal
for us to use MIDDMA_BYTE2 mode).
- opt: on tx: try and avoid flushing the internal buffer by padding out the
length of the last mbuf a bit (if possible)
- merged multiple DRQ/DTQ ADD macros into a single DRQ and a single DTQ
macro with a uniform interface to make the code simpler and easier to read.
- en_start: only update atm_flags if EN_MBUF_OPT is enabled (which it
should be)
- for alburst: make sure we don't DMA more bytes than we need (on both
tx and rx). if the alburst is larger than we need, drop to
MIDDMA_WORD mode.
- major change: enable the use of byte and 2 byte DMA on the trasmit side.
this allows us to DMA from non-word sized/aligned mbufs directly.
[the old code would always call en_mfix which would copy (or move) the
data in order to ensure proper alignment... it turns out TCP gives
us non-word sized/aligned mbufs when it is retransmitting, so we needed
to handle this case more efficiently.] the following functions
were changed to make this work:
- en_dqneed: add an arg to let us know if we are transmitting or not.
if we are TX, then we must take into account byte DMAs when
estimating the number of DTQs we will need for a buffer
- en_start: only mfix mbufs if DMA is disabled
- en_txdma: only set launch.nodma if we have en_mfix'd the mbuf chain
also, we may need a DTQ to flush the chip's internal byte buffer
- en_txlaunch: only attempt a copy if we have the proper alignment.
add byte dma code for the front and end of the buffer.
make sure the internal dma buffer is flushed out.
- stats: keep track of how many times we have to use byte sized DMA
midwayreg:
- add byte/2byte DMA defines
midwayvar:
- add new stat counter to monitor less-than-word lengthed DMA
start adding back in tracing printfs. add support for the virtual page
table. Now it gets to user-land code, but fails because i've not
added support to the context switch code to activate and deactivate pmaps.
more naturally in terms of way the OSF/1 PALcode delivers traps and
interrupts. Clean up fault/exception handling code and system entry
points. Seperate ASTs into a seperate C function.
(2) Add unaligned access fixup code to fix unaligned quad, long,
and IEEE S and T floating datum loads and stores. VAX floating data
types not yet supported, and in the future will only be supported if
FIX_UNALIGNED_VAX_FP is defined. (No point in wasting the space when
most of the time there will never be VAX FP loads and stores.) Right
now, these features can be controlled only by sysctl. The (boolean)
integer sysctls machdep.unaligned_print, machdep.unaligned_fix, and
machdep.unaligned_sigbus control printing about unaligned accesses
(defaults on), fixing up of unaligned accesses (defaults on), and
forcing a SIGBUS on unaligned accesses (defaults off). If an access
is not fixed up (for lack of method or explicit decision), a SIGBUS is
always generated to keep programs from using bogus data. At some point,
these three choices should be controlled by per-process flags, as well.
more naturally in terms of way the OSF/1 PALcode delivers traps and
interrupts. Clean up fault/exception handling code and system entry
points. Seperate ASTs into a seperate C function.
(2) Add a boot flag ('H' and 'h') to make sure the kernel never
reboots after panic. Useful for debugging kernels which panic early
on after user processes have started, to fend off infinite reboot cycles.
Sort boot flag switch.
(3) Add unaligned access fixup code to fix unaligned quad, long,
and IEEE S and T floating datum loads and stores. VAX floating data
types not yet supported, and in the future will only be supported if
FIX_UNALIGNED_VAX_FP is defined. (No point in wasting the space when
most of the time there will never be VAX FP loads and stores.) Right
now, these features can be controlled only by sysctl. The (boolean)
integer sysctls machdep.unaligned_print, machdep.unaligned_fix, and
machdep.unaligned_sigbus control printing about unaligned accesses
(defaults on), fixing up of unaligned accesses (defaults on), and
forcing a SIGBUS on unaligned accesses (defaults off). If an access
is not fixed up (for lack of method or explicit decision), a SIGBUS is
always generated to keep programs from using bogus data. At some point,
these three choices should be controlled by per-process flags, as well.
more naturally in terms of way the OSF/1 PALcode delivers traps and
interrupts. Clean up fault/exception handling code and system entry
points. Seperate ASTs into a seperate C function.
(2) Clean up interrupt handling slightly.
(3) Add machinery to make [fs]uswintr be a bit more careful when
determining if a fault was actually their fault. (Unfortunately, they're
not actually implemented now, anyway.)
(4) Minor cleanup.
handling based on information in the 2nd Ed. of the Alpha Architecture
Reference Manual.
(2) Clean up interrupt handling slightly.
(3) Clean up badaddr().
more naturally in terms of way the OSF/1 PALcode delivers traps and
interrupts. Clean up fault/exception handling code and system entry
points. Seperate ASTs into a seperate C function.
(2) Add machinery to make [fs]uswintr be a bit more careful when
determining if a fault was actually their fault. (Unfortunately, they're
not actually implemented now, anyway.)
and IEEE S and T floating datum loads and stores. VAX floating data
types not yet supported, and in the future will only be supported if
FIX_UNALIGNED_VAX_FP is defined. (No point in wasting the space when
most of the time there will never be VAX FP loads and stores.) Right
now, these features can be controlled only by sysctl. The (boolean)
integer sysctls machdep.unaligned_print, machdep.unaligned_fix, and
machdep.unaligned_sigbus control printing about unaligned accesses
(defaults on), fixing up of unaligned accesses (defaults on), and
forcing a SIGBUS on unaligned accesses (defaults off). If an access
is not fixed up (for lack of method or explicit decision), a SIGBUS is
always generated to keep programs from using bogus data. At some point,
these three choices should be controlled by per-process flags, as well.
more naturally in terms of way the OSF/1 PALcode delivers traps and
interrupts. Clean up fault/exception handling code and system entry
points. Seperate ASTs into a seperate C function.
(2) Restructure and improve machine check and correctable error
handling based on information in the 2nd Ed. of the Alpha Architecture
Reference Manual.
(3) Removed unused (and not likely useful) PALcode assembly stubs.
more naturally in terms of way the OSF/1 PALcode delivers traps and
interrupts. Clean up fault/exception handling code and system entry
points. Seperate ASTs into a seperate C function.
2441, and part of 2600. 2436 is left open to remind me that Chris's
comments are probably correct. This file should be built statically and
committed to the tree, not built on a per-kernel basis.
string with 0xFF instead of 0x20. Shorten the compare to 6 bytes to
take care of this.
Fixes PR #2617 by myself, which is based on a private problem report
by Neil Brewitt.
- Remove the kernel-jump-to-zero code as page zero mapping is removed
- Weed out cpu040 usage, use mmutype instead
- Fix doboot code
- add fill for page zero
- use m68k/sigreturn.s
- en_mfix shouldn't touch M_EXT mbufs. change to avoid this [at
the expense of allocating a new cluster mbuf to copy to].
XXX: en_mfix is getting called more often than I hoped it would
(seems to happen when TCP retransmits... we get all sorts of odd
sized, odd lengthed data mbufs, yuck). i may revise the xmit
DMA code to use non-word sized dma.
unnecessary, and given that there's no reason to waste time and
add complexity in trap handlers. This means that code which sets and
uses the USP has to be a bit more careful, but that doesn't happen much.
containing a substruct (the hardware frame) and an array of registers,
treat it like one big array of registers, for easier and prettier
access. Update everything to deal with that.
Process Control Block
Interrupt/Exception/Syscall Stack Frame
Machine Check Error Summary Register
Machine Check Logout Area
clean up some, and add prototypes for all of the CPU instruction and
PALcode function stubs.
- don't try and add a PDU trailer (or TBD for that matter) to an M_EXT
mbuf. when M_EXT is set, the data area of the mbuf can be shared
with other mbufs (see m_copym()), and writing to one M_EXT mbuf may
invalidate another M_EXT mbuf (saw this with PDU trailers).
- be sure to validate the length value from the PDU trailer so that it
doesn't throw us out of sync with the chip (fixes yet another source
of the dread "en_service: id mismatch" panic).
- on recv, don't bother allocating an mbuf if mlen is zero since we are
only going to toss the data anyway.
alpha_cpu.h, and this file is no longer necessary for the
Alpha. However, some machine-independent code (e.g.
/sys/compat/common/kern_exit_43.c) still wants a file with
this name to be around.
XID confusions with servers that cache them over a long period and
with clients that reboot quickly.
Problems: because of the sanity check that is done by comparing the system
time with filesystem time, XIDs will start at 0 until root is mounted,
which means it isn't completely safe for diskless setups. But it's clearly
better than it was. It would also be cleaner if all XID handling (more
generally, all RPC handling) within the kernel went through the
same functions.
matic media detection on Sun4m machines which support multiple media
(UTP and AUI). New function lenocarrier() switches media type when called
if no link? flags are set.
Redefine link flags for Lance ethernet (le):
<none>: switch between media as needed to find carrier (autodetect)
Preference to whatever the PROM is using (AUI if not booted
from the network)
link0: use UTP port only
link1: use AUI port only
This allows Sun4m machines with dual media Lance ethernet adapters to
boot from the network on the UTP port.
function is called when carrier is lost in lieu of a console message.
This is used by the sparc port to switch media types on Sun4m machines
if the appropriate link flags are set by ifconfig (see following commit
message). Should not affect ports that do not use sc_nocarrier().
we managed to kill the following SCSI-driver bugs for the Falcon:
- The Byte_Count_zero bit of the falcon DMA controller does not always
tell the truth! This caused the SCSI-driver to choke on devices that
disconnected in the middle of a DMA-transfer (mostly removables).
- Printing debug info about the 5380 on the Falcon is *only* permitted
when DMA is not active.
- Some functions forgot to remove possibly pending sofware interrupts
- Some debug options didn't allow debugging a single target while they
could easily be made to do so. This is fixed.
from the mach3 versions only as much as necessary to allow the old
NetBSD/Alpha pmap code to compile. THESE WILL NOT WORK AS-IS, and at
minimum will require code to implement reference- and modified-bit
emulation.
All Unibus dependencies are removed.
Can support different controllers on different buses.
Allows cloning of devices.
TODO:
Write support for MSCP tapes.
old config anymore, all devices should use new config by now.
Add structures to handle DMA devices using new config.
Fixed bug that caused unwanted ubareset's on 11/780.
Use new ldda() and stda() to make MultiCache initialization work on
SuperSparc-II machines (these machines require 64-bit access, even
to a 32-bit register).
existence of the SCSI hardware in the boards adress region.
Thanks to Oster Nerhus for the detailed bug report and to Ralph Schmidt for
the recommended probing method.
to ignore inbound data sizes less than the threshold. otherwise,
when connected to a video aal0 input the driver gives us a massive
stream of 56 byte mbufs each with one aal0 cell in it and the
system just can't keep up with it, especially if the socket buffer
size is large [it hangs until you turn off the video source].
fixes: when turning off a vc, try and check the freshest copy of the mode
when seeing if we need to enter the "drain" state. also, don't
panic if we get unexpected rx interrupt on a VCI (instead make sure
the VC is off, print a warning, and move on!).
- change rxso into a void handle ... no need to know details at this low
level.
fix:
- forgot to init "vci" during ENOTHER_DRAIN case of RX DMA interrupt
[could cause a vm_fault on native mode connection close depending
on needing to drain out the VC and also the random value of vci]
- more stat counters (rxmbufout -- mbufs, ttrash -- trashed RBD)
- add swsl_size to keep track of size of software slist
- revise when we call en_service
- fix handling of circular lists (was getting full vs empty confused in
extreem cases...)
- native mode atm: remove sbappend code: can't do it here since
socket buffer is protected only by splsoftnet and that doesn't
mask enintr off. forces us to switch to a two level interrupt
scheme. revise call to atm_input to reflect that.
- Good bye SYS_PBOOT and SYS_NBOOT, hello SYS_UBOOT (unified boot)
Currently supports booting from SCSI and HP-IB disk and network
from a single boot block. Infrastructure for booting from
HP-IB tape is there, but it doesn't quite work yet.
- Add a slightly modified version of Gordon Ross's "rawfs",
which provides a filesystem-like interface to tape devices.
Still needs debugging (see above).
- Rename sys_inst.c to inst.c, so that the LIF directory entry
turns out right (used to look like SYS_SYS_I if loaded from
tape).
- Add a "netio_ask" hint to netio.c, so that a special instnetio.o
doesn't have to be compiled for SYS_INST. Defaults to using
bootparams, but if set, will prompt user for information
usually obtained from bootparams.
- General cleanup.
in a more elegant way, but it works as is, and we need it now.
- Add a README and Makefile for the above. Note that this directory should
NOT be added to ../Makefile's SUBDIRS.
somehow got lost between NetBSD and Linux. Output side mbuf-walking
code now correctly bytestuffs mbuf chains, as well as single mbufs
and clusters.
* Update radio error-parsing code to Stuart's latest stable code.
We now parse error messages the older code didn't.
* Note where radio-crash watchdogs should be added (the linux code
is there, #ifdef'ed out). This still just doesn't work.
The radio reset doesn't always work even when slattatch is first started,
but I have a radio with old firmware, which may be a contributing factor.
* Correct the checks for the tty output queue being overfull; estimate
the stuffed pkt size as (original * 65/64) + STRIP_HDRLEN + 2,
instead of SLIP's (2*SLMTU). Re-enable the disabled check now the
size estimate isn't excessively large.
* Fix BPF tapping of strip interfaces, STRIP packets are wrapped
in a SLIP bpf header. This implies no BPF support for arp or atalk,
even though Linux boxes are sending arp requests and gratuitous arps.
There may be no good fix short of adding explicit STRIP encapsulation
support to bpf/tcpdump.
* Still need a solid walkthrough, and rewrite to eliminate redundant
receive-side mbuf copying.
[1] if user tries to enter in a bogus PVC don't leave it in the routing
table ... remove it
[2] change ioctl arg to include rxso for lower layer
[3] add hooks (inside "NATM" ifdef) for native mode atm sockets so that
they don't clash with IP PVCs. [i am still debugging the native
mode atm socket protosw code]
[1] rxslot "cur" pointer wasn't updated if we copied all data and
didn't have any fill. this could cause a service id mismatch
panic under load.
[2] recv of aal0 forgot to include the cell header in the data length
calc [add cell header length to midwayreg.h]
[3] fix control info "chop" for raw mode and aal0
[4] handle rxsb better for native atm connections
(XXX: still working on this ... not complete and needs revision)
- ability to be either a BROADCAST or POINTTOPOINT interface.
- a humble beginning of link-layer addressing (differs from PR
by using a `struct sockaddr' instead of single byte).
- Use an extent map to manage the ISA memory "hole", much like
how i/o port space is manged with an extent map. Do the actual
accounting in bus_mem_{,un}map().
- When creating the ioport and iomem extent maps, pass the
EX_NOBLOB flag, which tells the extent map code to disallow
fragmenting of regions. This will make it easier to catch
simple i/o and memory mapping bugs (like the if_ed.c bug I just
fixed) in the future.
- In bus_mem_map(), a chunk of code was indented using spaces.
Make the indentation use tabs.
- In bus_io_unmap(), fix the format passed to printf().
physical memory is sized by a loop that writes data to the first
word in a page, (writes something else to settle the bus) and then reads
back the word it wrote. If the read succeeds, the amount of physical
memory is increased by one page.
This fails on a 5000/1xx with a memory subsystem filled with 8 low-density
(4Mbyte) SIMMs. The memory-decoding hardware aliases the 32Mbytes of
physical memory at physical addresses 0, and at 32M (and presumably
at 64 and 96Mbytes.) The contiguous aliasing causes the memory-sizing
loop to continue at 32 MBytes, testing the memory that's really
at address 0, overwriting and crashing the kernel.
Fixed (for 1.2) by reading the SIMM-decoder stride size from the
motherboard, and reducing the loop bound to 32Mbytes on a 5000/1xx
with low-density SIMMs. (Other models have a non-power-of-2 maximum
memory and so are not subject to _contigous_ aliasing of physical memory).
This takes care of two related problems:
- `umount -f' wouldn't work if someone's working directory is
the filesystem root.
- vfs_unmountall() would complain about a busy `/' on a
diskless setup.
The physical memory-sizer claims to preserve memory contents
(specifically the contents of msgbuf). The loop writes different
values into two adjacent locations and reads the contents of the
first, to ensure that whatever is read back from the first location is
from memory and isn't just the first write persisting on the bus.
The loop preserved the value of the first location, but not the second,
resulting in the second test value ('ZZZZ') over-writing a word in msgbuf.
- A fixed extent map (statically allocated descriptor storage) is
created in init386(), just before the call to consinit(). The
fixed descriptor storage has enough room for 8 region entires,
which is plenty for early initialization, but doesn't chew up
that much memory.
This extent map (ioport_ex) manages the i386 i/o port
space (0x0 - 0xffff).
- Just before the call to configure() in cpu_startup(), a
flag is set which notifies the bus_io functions that it is
safe to use malloc() to allocate descriptor storage, in the
event that more than 8 regions are needed.
- bus_io_map() attempts to allocate the specified region from
ioport_ex. If the allocation succeeds, the io handle is
filled in. If the allocation fails, it is implied that
something else is already using that io space, and an
error condition is returned.
- bus_io_unmap() frees a region previously allocated from
ioport_ex in bus_io_map(). If the free fails, a warning
is printed on the conole.
These changes implement "port accounting". This is required for
proper autoconfiguration on the i386 port, and makes dealing with,
among other things, PCMCIA io mappings _much_ easier.
Understands allocation aligment and boundary restrictions, "specific region"
allocations, and suballocations. Capable of statically or dynamically
allocating map overhead.
Many thanks to Matthias Drochner for running the code for me, and sending
me bug fixes, optimizations, and suggestions. Also, many thanks to
Chris Demetriou for his extremely helpful suggestions.
XXX No manual page yet. One is forthcoming, as soon as I can scare up
the time to write one. This has been sitting on my plate for quite a
while, and several projects are waiting for it. Time to move on.
piece of hardware into kernel virtual memory (was only a guess 'til now).
XXX Unfortunately, the hardware vendor reserves the right to change this in
future DraCo revisions. We must rethink delay() and DELAY(), at least for the
DraCo, soon.
which was taken from OpenBSD/pica.
The previous revision of elf.c replaced Ted Lemon's elf exec machinery
with something closer to Christos' MI elf machinery. It turns out
that old NetBSD/pmax elf binaries have three segments, and the newer
elf exec machinery cannot exec them.
The old elf exec machinery is folded back into cpu_exec.c, which falls
back onto using the old machinery if the new machinery fails. The
old-style binaries will be deprecated at the 1.2 release.
is close to how I reworked it, I pulled in the (essentially MI) NetBSD/alpha
and added the appropriate support around it.
- No need for the GENERIC kernel config option.
- NFS-mounted root and swap are supported.
- If we can't figure out where the root filesystem is from what the Booter
tells us, ask the user.
- Split the mainbus autoconfig code to a separate file.
Also, update/add copyrights as appropriate.
delayed write is logically converted to a sync write, mirroring the async case.
In bdwrite(), move the tape case earlier to avoid needless reassignbuf()s.
where the console is, because that is correct if, for example, you
boot with the diag/norm switch set for diagnostics and the EEPROM
is set for keyboard/display.
* Update arch/mips/mips/cpu_exec.c to include MI exec_elf.h header,
and to use the MI interface exec_elf_makecmds().
* Replace arch/mips/mips/elf.c (Ted Lemon's elf code) with
a version of Christos's MI elf exec code, munged to support demand paging
and mips shared libraries.
Show proper inquiry data when version >= 1, not <= 1 (SCSI-2 devices
now display their inquiry infor).
Correct transfer length passed to dk_unbusy (b_resid hasn't been
updated yet).
* Eliminate the #ifdef DS5000 and kernel config option DS5000 altogether.
option DS5000 was a hangover from 4.4bsd/pmax; it's never been clear
if it means "support for DS5000/200" or "support for any TC machine".
This implictly fixes pr 1828.
* Use the "tc.h" and NTC generated by config for a "tc? at mainbus"
where appropriate, intsead of #ifdef DS5000.
* Canonicalize the spelling of kernel options that enable support
for particular models. Use
DS5000_25 for the Personal Decstation aka MAXINE aka KN02-ca
DS5000_100 for the 5000/1xx series aka KMIN aka KN02-ba
DS5000_200 for the 5000/200 aka 3MAX aka KN02
DS500_240 for the 5000/240 aka 3MAXPLUS aka KN03
and change DS_5000_xxx to DS5000_xxx everywhere.
* Wrap the interrupt handlers for each model in the appropriate #ifdef,
instead of wrapping all of them with "#ifdef DS5000".
* Wrap the TC autoconfig for each model in the appropriate #ifdef.
* Eliminate the #ifdef DS5000 and kernel config option DS5000 altogether.
option DS5000 was a hangover from 4.4bsd/pmax; it's never been clear
if it means "support for DS5000/200" or "support for any TC machine".
This implictly fixes pr 1828.
* Use the "tc.h" and NTC generated by config for a "tc? at mainbus"
where appropriate, intsead of #ifdef DS5000.
* Canonicalize the spelling of kernel options that enable support
for particular models. Use
DS5000_25 for the Personal Decstation aka MAXINE aka KN02-ca
DS5000_100 for the 5000/1xx series aka KMIN aka KN02-ba
DS5000_200 for the 5000/200 aka 3MAX aka KN02
DS500_240 for the 5000/240 aka 3MAXPLUS aka KN03
and change DS_5000_xxx to DS5000_xxx everywhere.
* Wrap the interrupt handlers for each model in the appropriate #ifdef,
instead of wrapping all of them with "#ifdef DS5000".
