o Fix the packet statistics
o Make sure we set the FD bit when in full duplex
o Improve TX side efficency by eliminating a data copy for
unfragmented mbufs (the hardware can't do s/g).
o Minor busdma pedantry
o better comments in some places, more XXX in others
o Minor style nits.
This solves a problem I was seeing where I'd get no ethernet when not
booting with a NFS root. Well, unless I unplugged the cable and
plugged it back in first so I'd get the same up down up messages I get
for NFS root...
Thanks to sam and scottl for suggestions on making this driver more
efficient through better use of approrpiate APIs.
a lock to prevent interspersed strings written from different CPUs
at the same time.
To avoid putting a buffer on the stack or having to malloc one,
space is incorporated in the per-cpu structure. The buffer
size if 128 bytes; chosen because it's the next power of 2 size
up from 80 characters.
String writes to the console are buffered up the end of the line
or until the buffer fills. Then the buffer is flushed to all
console devices.
Existing low level console output via cnputc() is unaffected by
this change. ithread calls to log() are also unaffected to avoid
blocking those threads.
A minor change to the behaviour in a panic situation is that
console output will still be buffered, but won't be written to
a tty as before. This should prevent interspersed panic output
as a number of CPUs panic before we end up single threaded
running ddb.
Reviewed by: scottl, jhb
MFC after: 2 weeks
The 'nooption' kernel config entry has to be used to turn KSE off now.
This isn't my preferred way of dealing with this, but I'll defer to
scottl's experience with the io/mem kernel option change and the grief
experienced over that.
Submitted by: scottl@
syscalls using __syscall but only actually returning 32bits, such as mmap(),
specially : they set the return value in td->td_retval[0], but the userland
functions will expect this in r1, and not in r0 as it is normally done, as it
is the LSB. So add a special case for all these syscalls (all except lseek,
which truly returns 64bits).
Many thanks to Peter Grehan for his patience while explaining me the issue.
This interface also appears in the AT91SAM9260 and '61 as well as the
AVR32 based micros from Atmel. We don't yet support write protect or
hot-swap in this bridge driver.
Submitted by:
Reviewed by:
Approved by:
Obtained from:
MFC after:
Security:
Move the relocation definitions to the common elf header so that DTrace
can use them on one architecture targeted to a different one.
Add the additional ELF types defines in Sun's "Linker and Libraries"
manual.
KERNBASE and VM_MAXUSER_ADDRESS.
Remove the useless include of opt_global.h, as noticed by netchild@ (the one
in arm/elf_trampoline.c is legit, because this file is compiled outside the
kernel, and doesn't use the standard CFLAGS).
whole the physical memory, cached, using 1MB section mappings. This reduces
the address space available for user processes a bit, but given the amount of
memory a typical arm machine has, it is not (yet) a big issue.
It then provides a uma_small_alloc() that works as it does for architectures
which have a direct mapping.
Originally, I had adopted sparc64's name, pmap_clear_write(), for the
function that is now pmap_remove_write(). However, this function is more
like pmap_remove_all() than like pmap_clear_modify() or
pmap_clear_reference(), hence, the name change.
The higher-level rationale behind this change is described in
src/sys/amd64/amd64/pmap.c revision 1.567. The short version is that I'm
trying to clean up and fix our support for execute access.
Reviewed by: marcel@ (ia64)
mark system calls as being MPSAFE:
- Stop conditionally acquiring Giant around system call invocations.
- Remove all of the 'M' prefixes from the master system call files.
- Remove support for the 'M' prefix from the script that generates the
syscall-related files from the master system call files.
- Don't explicitly set SYF_MPSAFE when registering nfssvc.
implementations and adjust some of the checks while I'm here:
- Add a new check to make sure we don't return from a syscall in a critical
section.
- Add a new explicit check before userret() to make sure we don't return
with any locks held. The advantage here is that we can include the
syscall number and name in syscall() whereas that info is not available
in userret().
- Drop the mtx_assert()'s of sched_lock and Giant. They are replaced by
the more general checks just added.
MFC after: 2 weeks
The core uart code expects the receive method to actually puts the
characters read into its buffers. For AT91, it's done in the ipend routine,
so also check if we have the alternate break sequence here.
MFC after: 3 days
Introduce framework to configure the multiplexed pins on boot.
Since the USART supprots RS-485 multidrop mode, it allows the
TX pins to float. However, for RS-232 operations, we don't
want these pins to float. Instead, they should be pulled up
to avoid mismatches. Linux does something similar when it
configures the TX lines. This implies that we also allow the
RX lines to float rather than be in the state they are left in
by the boot loader. Since they are input pins, I think that
this is the right thing to do.
Plus minor for our board.
install custom pager functions didn't actually happen in practice (they
all just used the simple pager and passed in a local quit pointer). So,
just hardcode the simple pager as the only pager and make it set a global
db_pager_quit flag that db commands can check when the user hits 'q' (or a
suitable variant) at the pager prompt. Also, now that it's easy to do so,
enable paging by default for all ddb commands. Any command that wishes to
honor the quit flag can do so by checking db_pager_quit. Note that the
pager can also be effectively disabled by setting $lines to 0.
Other fixes:
- 'show idt' on i386 and pc98 now actually checks the quit flag and
terminates early.
- 'show intr' now actually checks the quit flag and terminates early.
address is in the userland address space. The proper thing is either to choose
a virtual address in the kernel address space beyond the KVA, or to use
pmap_mapdev().
we're unable to allocate the memory for a PTE, we'll wait until we can. If not,
we'll just return.
Use M_NOWAIT|M_USE_RESERVE to allocate PTEs, it is less aggressive than
M_NOWAIT alone.
Suggested by: alc
Make serial ports more robust and reliable. Make non-console ports
work. This might have broken skyeye stuff.
o Introduce ping-pong receive buffers.
o Use DMA to copy characters directly into memory.
o Support baud rates other than 115200
o Use 1 stop bit when 1 stop bit is requested (otherwise 2 were used,
which caused dropped characters when received in bursts).
o Use 1.5 stop bits for 5-bit bytes, and 2 stop bits otherwise when 2
stop bits were requested.
o Actually update line parameters.
o Fix comments
o Move init into attach
o Tweaks to TX interrupt registers to get them reliable and non-storming.
o harvest data in ipend since the latency between it and the callback
was too long. This likely is how it should be, I don't know why I deferred
things to the callback before.
o disable all interrupts in console init. We don't want interrupts until
we turn on an ISR.
o cosmetic tweaks
o Automatically detect of the TIMEOUT interrupt is supported. If so, use
it so we get better CPU utilization. Otherwise do a character at a time
RX. Good news here is that it seems we have enough CPU and low enough
fast interrupt latency to do this reliably.
o Don't read USART_CR. It is a write-only register.
o start to implement bus_ioctl. Do BAUD now...
number of banks, rows and columns the SDRAMC is programmed to access
to determine the RAM size for the board, rather than hard-wiring it to
be 32MB. My company's board with 64MB now probes correctly, as does
the KB9202 with only 32MB. This means that to detect the right memory
size, our boot loader must correctly initialize these values. This is
a fairly safe assumption because the boot loader has to initialize
SDRAM already, and it isn't really possible to change this register
after we've accessed SDRAM.
The boot loader is supposed to leave this bit set to the right value
for the board. If this bit was set at attach time, use it to init the
config register correctly.
Note: this means the boot loader has to properly initialize it.
an explicit comment that it's needed for the linuxolator. This is not the
case anymore. For all other architectures there was only a "KEEP THIS".
I'm (and other people too) running a COMPAT_43-less kernel since it's not
necessary anymore for the linuxolator. Roman is running such a kernel for a
for longer time. No problems so far. And I doubt other (newer than ia32
or alpha) architectures really depend on it.
This may result in a small performance increase for some workloads.
If the removal of COMPAT_43 results in a not working program, please
recompile it and all dependencies and try again before reporting a
problem.
The only place where COMPAT_43 is needed (as in: does not compile without
it) is in the (outdated/not usable since too old) svr4 code.
Note: this does not remove the COMPAT_43TTY option.
Nagging by: rdivacky
There is a race with the current locking scheme and removing
it should have no measurable performance impact.
This fixes page faults leading to panics in pmap_enter_quick_locked()
on amd64/i386.
Reviewed by: alc,jhb,peter,ps
- Try hard to calculate a safe sp, so that the stack doesn't get smashed
while uncompressing or relocating the kernel.
- Bring in code needed to calculate the cacheline size etc, needed for
arm9_idcache_wbinv_all.
current interface with the machine-independent layer. Without this change,
the page daemon would only have been awakened the first time that the
number of pv entries went above the high water mark, not each time.
the first and last cache line in PREREAD, and just invalidate the cache
lines in POSTREAD, instead of write-back/invalidating in POSTREAD, which
could lead to stale data overriding what has been transfered by DMA.
enabled. It has been commented out for a reason I forgot but I suspect
does not apply anymore.
Technically speaking it's not required to do it, has the data and the
instruction cache have been disabled in _start(). However, it may change
in the future, so I don't want to rely on this behavior.
Submitted by: kevlo
Add a new option, SKYEYE_WORKAROUNDS, which as the name suggests adds
workarounds for things skyeye doesn't simulate. Specifically :
- Use USART0 instead of DBGU as the console, make it not use DMA, and manually provoke an interrupt when we're done in the transmit function.
- Skyeye maintains an internal counter for clock, but apparently there's
no way to access it, so hack the timecounter code to return a value which
is increased at every clock interrupts. This is gross, but I didn't find a
better way to implement timecounters without hacking Skyeye to get the
counter value.
- Force the write-back of PTEs once we're done writing them, even if they
are supposed to be write-through. I don't know why I have to do that.
vm_ksubmap_init() calls pmap_copy_page(), which uses the mini data cache
to do the copy, but we're running uncaching before cpu_setup().
For some reason it hasn't been a problem so far, but it is for the
PXA255.
Spotted out by: benno
date: 2006/04/12 04:22:50; author: alc; state: Exp; lines: +14 -41
Retire pmap_track_modified(). We no longer need it because we do not
create managed mappings within the clean submap. To prevent regressions,
add assertions blocking the creation of managed mappings within the clean
submap.
Reviewed by: tegge
the wire. This increases the speed considerably. Start to put
infrastructure in place to do RX side, but that requires more study
before it can be done.
relocate it), do not attempt to call pmap_vac_me_harder() on the page.
At this point m will be NULL, and we know we won't have any cache
issues with this page.
Add bus attachment for the ohci device on this chip. The bus and hub
are detected correctly, but the children devices aren't detected
correctly for reasons unknown.
o update TODO list
o Better use of busdma
o mark RX dtors as COHERENT. This helps performance a lot by not requiring
so many EXPENSIVE cache flushes. The cost of accessing it non-cached
is much smaller.
o Copy data from Rx buffers to make IP header 4 byte aligned.
o CRC length included in reported length, so cope
o Don't free TX buffer twice
o Manage TX buffers better.
o Enable just the interrupts we want.
o Manage OACTIVE better
# Some of these done by cognet
# These changes let us get to # via NFS root.
o Add memory barrier to bus space
o Allow for up to 3 IRQs per device
o Move to table driven population of children devices.
o Add support for usb ohci memory mapped controller resource allocation.
o Clean up a bunch of extra writes to disable interrupts that are now
done elsewhere.
o Force all system interrupt handlers be fast. We get deadlock if they
aren't.
o Disable all interrupts that the ST can generate until we have an ISR
to service them.
o Correct clock calculation to make DELAY the right length...
Submitted by: cognet (#2)
with malloc() or contigmalloc() as usual, but try to re-map the allocated
memory into a VA outside the KVA, non-cached, thus making the calls to
bus_dmamap_sync() for these buffers useless.
- Reorder the events in exit(2) slightly so that we trigger the S_EXIT
stop event earlier. After we have signalled that, we set P_WEXIT and
then wait for any processes with a hold on the vmspace via PHOLD to
release it. PHOLD now KASSERT()'s that P_WEXIT is clear when it is
invoked, and PRELE now does a wakeup if P_WEXIT is set and p_lock drops
to zero.
- Change proc_rwmem() to require that the processing read from has its
vmspace held via PHOLD by the caller and get rid of all the junk to
screw around with the vmspace reference count as we no longer need it.
- In ptrace() and pseudofs(), treat a process with P_WEXIT set as if it
doesn't exist.
- Only do one PHOLD in kern_ptrace() now, and do it earlier so it covers
FIX_SSTEP() (since on alpha at least this can end up calling proc_rwmem()
to clear an earlier single-step simualted via a breakpoint). We only
do one to avoid races. Also, by making the EINVAL error for unknown
requests be part of the default: case in the switch, the various
switch cases can now just break out to return which removes a _lot_ of
duplicated PRELE and proc unlocks, etc. Also, it fixes at least one bug
where a LWP ptrace command could return EINVAL with the proc lock still
held.
- Changed the locking for ptrace_single_step(), ptrace_set_pc(), and
ptrace_clear_single_step() to always be called with the proc lock
held (it was a mixed bag previously). Alpha and arm have to drop
the lock while the mess around with breakpoints, but other archs
avoid extra lock release/acquires in ptrace(). I did have to fix a
couple of other consumers in kern_kse and a few other places to
hold the proc lock and PHOLD.
Tested by: ps (1 mostly, but some bits of 2-4 as well)
MFC after: 1 week
Rename struct thread's td_sticks to td_pticks, we will need the
other name for more appropriately named use shortly. Reduce it
from uint64_t to u_int.
Clear td_pticks whenever we enter the kernel instead of recording
its value as reference for userret(). Use the absolute value of
td->pticks in userret() and eliminate third argument.
hack where it assumes the first field of the driver softc is the struct
ifnet, and it copies its value in mii_phy_probe().
- In the interrupt handler, set the mbuf m_len field on packet receive.
