I'm at it also add a comment in mtx_validate() explaining the purpose
of the last change.
Basically, this fixes booting kernels compiled with MUTEX_DEBUG. What used
to happen is before we setidt from init386() [still using BTX idt], we
called mtx_init() on several mutex locks, notably Giant and some others.
This is a problem for MUTEX_DEBUG because it enables mtx_validate() which
calls kernacc(), some of which in turn requires Giant.
Fix by calling kernacc() from mtx_validate() only if (!cold).
explain the subtle side effects that are going to happen and why we go
ahead and ack the interrupt source. This stuff is tricky to get
right.
Also, emperical tests have shown that doing a shutdown in attach to be
ineffectual, so remove it from there. Analysis of the code paths
shows that nearly identical writes to these registers happen in later
parts of the code. The hanging problem on thinkpads when we change
the interrupt routing type is something else.
shutdown and also before we get going with the device initialization.
This may fix the hangs some people are seeing on warmboot. It appears
that some machines will reset the cardbus bridge on boot, while others
don't. So we turn off the card, and ack the interrupts (which likely
is a nop in the shutdown case since we're still fielding interrupts).
This should turn off the interrupts.
Since I don't have hardware that hangs on reboot, I'm committing this
without testing that aspect of the patch (it causes no harm on my
Dell).
initialization structure. Warn the user for those chipsets that
aren't yet customized that they might not work. Second, try to power
off the slot on attach and ack the interrupts. I don't know, but this
might solve the hangs that people will see on Thinkpads if they set
hw.pcic.init_routing=1.
go into the PLX 9052's interrupt control register and turn on the magic
interrupt enable bit.
Partial thanks are due to OpenBSD for pointing out that the chip is a
PLX 9052 and pointing me to the datasheet PDF.
then one can restart from a panic by resetting the panicstr variable to
NULL. This commit conditionalizes the previously committed functionality
on this variable. It also removes the __dead2 attribute from the panic()
function so that when one continues from a panic() the behavior will
be predictable.
* add new channels to the end of the list so channels used in order of
addition
* de-globalise definition of struct snddev_info and provide accessor
functions where necessary.
* move the $FreeBSD$ tag in each .c file into a macro and allow the
/dev/sndstat handler to display these when set to maximum verbosity to aid
debugging.
* allow each device to register its own sndstat handler to reduce the amount
of groping sndstat must do in foreign structs.
It appears that a number of PrismII card vendors seem to be doing the
same thing (that is, using the same PCI bridge chip) to support PCI,
but each with their own vendor/product ID. So rather than cut-n-paste
another if statement into the probe routine, it's probably better to
provide support for a table. Adding new devices will be a lot easier
that way.
sysproto.h in addition to the existing padding afterwards.
This is needed to support big-endian architectures like sparc64.
Reviewed by: bde
Tested on alpha by: jhb
timeout callwheel and buffer cache, out of the platform specific areas
and into the machine independant area. i386 and alpha adjusted here.
Other cpus can be fixed piecemeal.
Reviewed by: freebsd-smp, jake
that the directory entry was in use before attempting to find it
in the hash structures to change its offset. Normally, unused
entries do not need to be moved, but fsck can leave behind some
unused entries that do. A dirhash sanity panic resulted when the
entry to be moved was not found. Add a check that stops entries
with d_ino == 0 from being passed to ufsdirhash_move().
In order to ensure security and functionality, RFC 1948 style
initial sequence number generation has been implemented. Barring
any major crypographic breakthroughs, this algorithm should be
unbreakable. In addition, the problems with TIME_WAIT recycling
which affect our currently used algorithm are not present.
Reviewed by: jesper
purely informational and can give some advance indications of tuning
problems. These are i386 only for now as it seems that the i386 is
the only one suffering kvm pressure.
trap_fatal() to make restarting from panic's slightly easier. Before if
one did 'w 0 0' in ddb, the longjmp in ddb inside of trap_fatal() would
result in Giant being held (or recursed one level deeper) which led to
problems later on. You can now drop to teh debugger, do 'w 0 0', and
continue w/o a problem.
callout_stop() would fail in two cases:
1) The timeout was currently executing, and
2) The timeout had already executed.
We only needed to work around the race for 1). We caught some instances
of 2) via the PS_TIMEOUT flag, however, if endtsleep() fired after the
process had been woken up but before it had resumed execution,
PS_TIMEOUT would not be set, but callout_stop() would fail, so we
would block the process until endtsleep() resumed it. Except that
endtsleep() had already run and couldn't resume it. This adds a new flag
PS_TIMOFAIL to indicate the case of 2) when PS_TIMEOUT isn't set.
- Implement this race fix for condition variables as well.
Tested by: sos
request and just calling it when we get a bridge interrupt. The
problem is that if other code wants to block hardware interrupts for a
little bit with splXXX, those masks aren't updated the way we're doing
it. This doesn't matter for -current, but does for -stable.
The whole reason that we were catching interrupts was to detect that
the card was still there. Ian's fixes however ensure that the card
will be there with an interrupt handler, or not there at all. Since
the pcic interrupt is at a high priority, this should be OK.
This should fix the network related crashes people started seeing in
stable after I merged the pcic as a pci device code.
Submitted indirectly by: Ian Dowse
MFC when: Ian has had a chance to do his torture hang testing.
Add tagged queueing support for new IBM drives.
Add support for Yet Another Promise ATA 100 chip.
Flush disk cache on close.
Dont flush the disk cache on BIO_ORDERED anymore.
Cleanup the tests for DMA on ATAPI devices.
Allow to share ALL irq's even the std irg 14 & 15.
Fix calculation bug in end of media code on CD's.
Add REZERO on opening a CDR/CDRW.
Cleanup ataioctl a bit.
and such was just a bad idea and one that users should be forced to
enable if they want it. This patch introduces a hw.pci.enable_pcibios
tunable for those people. This does not impact the pcibios interrupt
routing at all.
Approved by: peter, msmith
some bios vendors took it apon themselves to "censor" the
host->pci bridges from PCIBIOS callers, even when the caller
explicitly asks for them. This includes certain Compaq machines
(eg: DL360) and some laptops.
If we detect this, shut down pcibios and revert to using IO
port bashing.
Under -current, apcica does a better job anyway.
during trap handlers.
Implement ptrace_set_pc FWIW.
Initialize the pcb window scratch area in setregs(), and setup user
registers as specified by the SCD.
Submitted by: tmm
kernel from usermode. The remaining user windows are spilled
to the pcb as necessary. The user land window fault handlers
fill directly from the pcb on return.
Add system call entry points.
Submitted by: tmm
has existed for a long time, but I made it worse a few months ago
by by adding calls to VFS_ROOT() and checkdirs() in revision 1.179.
Also, remove the LK_REENABLE flag in the lockmgr() call; this flag
has been ignored by the lockmgr code for 4 years. This was the only
remaining mention of it apart from its definition.
Reviewed by: jhb
that do not have valid NVRAM. In particular, we were leaving
a retry count set (to retry selection timeouts) when thats
not really what we want. Do some constant string additions
so that LOGDEBUG0 info is useful across all cards.
MFC after: 2 weeks
information. The default limits only effect machines with > 1GB of ram
and can be overriden with two new kernel conf variables VM_SWZONE_SIZE_MAX
and VM_BCACHE_SIZE_MAX, or with loader variables kern.maxswzone and
kern.maxbcache. This has the effect of leaving more KVM available for
sizing NMBCLUSTERS and 'maxusers' and should avoid tripups where a sysad
adds memory to a machine and then sees the kernel panic on boot due to
running out of KVM.
Also change the default swap-meta auto-sizing calculation to allocate half
of what it was previously allocating. The prior defaults were way too high.
Note that we cannot afford to run out of swap-meta structures so we still
stay somewhat conservative here.
the ISR. We keep track of the card state and don't call the IRS when
the card isn't inserted. This helps quite a bit with card ejection
problems that Ian was seeing.
Submitted by: Ian Dowse
MFC upon: re approvel.
this allows constants to be copied without giving errors.
(removes a couple of warning messages in the kernel.)
Note: uipc_mbuf.c is a forced commit as it was accidentally included in another commit.
MFC after: 2 weeks
`struct xucred` with the credentials of the connected peer.
Obviously this only works (and makes sense) on SOCK_STREAM
sockets. This works for both the connect(2) and listen(2)
callers.
There is precise documentation of the semantics in unix(4).
Reviewed by: dwmalone (eyeballed)
should be initialized to 0 for PCI adapters which was a side-effect of
the memset in ahc_alloc(). EISA/VL adapters require our unpause value to
include the "levelness" of our interrupt.
This change allows EISA adapters to work again.
PR: 29689, 29471
prevent/workaround TX lockups in this driver. The secret seems to be to
not let the TX DMA queue become too full. If we have too many packets
in the queue, we should wait for them to drain a bit before trying to
queue more. This should prevent the lockup from occurring, and if it
does occur, there is special code in sf_start() to kick the NIC in the
head and get it going again.
Special thanks to Glen Neff for helping me test this fix.
to properly clear the interrupt register on the no error case. Also,
set the mcr register to zero when we find we can't support the chip.
This fixes the hang on sio driver attach problem in the new pci pccard
code that some people have reported. At least on my machine. I'd
like to get this into 4.4.
Submitted by: bde
PR: kern/29742
MFC after: 1 day
unnecessary breakage.
While here, use explicit sizes for the string fields so that we dont
have unintentional changes again in the future when key tunables change.
This still is not quite right, but a june userland is happy with
a -current kernel with these tweaks.
probably harmless in this case, since the latter is called on tty
input, which is usually a result of some system call, so we've got
plenty of stack left. It's still nice to fix these things, though, in
case somebody ever decides this driver is a good example of something
(perhaps "what you probably shouldn't do").
night of testing before merging to -stable.
Also added to code to detect TX underruns and automatically increase the
TX threshold to avoid them. Carefully placed diagnostig printf() about
this under #ifdef DIAGNOSTIC to avoid getting any panicky e-mails from
confused users, like I always do with the xl and dc drivers.
calling vtophys() and contigmalloc()/contigfree() directly. Hopefully,
I have shaken out all of the problems with busdma on the alpha now.
(Everything seems to work as expected.)
Also, change the max RX DMA limit to 1024 bytes instead of "unlimited,"
as the latter seems not to work correctly on the alpha that I tested.
(At 100Mbps, all attempts to receive frames yield RX errors.)
but it's better than the buggy behavior we have now. If we contigmalloc()
buffers in bus_dmamem_alloc(), then we must configfree() them in
bus_dmamem_free(). Trying to free() them is wrong, and will cause
a panic (at least, it does on the alpha.)
I tripped over this when trying to kldunload my busdma-ified if_rl
driver.
label if the dump device overflaps the label (which is a slight
misconfiguration). Dump routines don't use dscheck(), so the normal
write protection of the label doesn't help.
Reduced some nearby overflow bugs. In disk_dumpcheck(), there was
(fatal but fail-safe) overflow on i386's with 4GB of memory, at least
if Maxmem was the top page (can this happen?). The fix assumes that
the sector size divides PAGE_SIZE (dump routines already assume this).
In setdumpdev(), the corresponding overflow occurred with only about
2GB of memory on all machines with 32-bit ints. This allowed setdumpdev()
to succeed when it shouldn't have, but then disk_dumpcheck() failed
safe later. Except in old versions of FreeBSD like RELENG_3 where
there is no disk_dumpcheck().
PR: 28164 (label clobbering part)
MFC after: 1 week
to see if it was malloc()ed first" bug. In bus_dmamap_create(), one of
two things can happen: either we need to allocate a special map due to
some manner of bounce buffering requirement, or we can DMA a buffer
in place. On the x86 platform, the "in place" case results in
bus_dmamap_create() returning a dmamap of NULL. The bus_dmamap_destroy()
routine later checks for NULL and won't bother free()ing the map if
it detects this condition.
But on the alpha, we don't use NULL, we use a statically allocated map
called nobounce_dmamap(). Unfortunately, bus_dmamap_destroy() does not
handle the condition where we attempt to destroy such a map: it tries
to free() the dmamap, which causes a panic.
Fix: test that map != &nobounce_dmamap before trying to free() it.
With this fix, my busdma-ified if_sis driver works on the alpha. I'm
a bit alarmed that I'm the first person ever to trip over this bug, since
we have been using busdma on the alpha for a while, and since it sort
of screams out "Hi! I'm a bug! Booga-booga!" when you look at it.
(Somewhere, somebody will say: "But Bill, why don't you just not bother
destroying the maps in this case." Because the API is supposed to be
a) symetrical and b) opaque to the caller. I can't know whether it's safe
to skip the bus_dmamap_destroy() step or not without sticking my fingers
into unsafe places, which is what I wanted to avoid in the first place.)
assing an IRQ. Add better comments while I'm here.
MFC after: 1 day
# Note: That's merging all the -current pci pcic code, not just this one
# change for the Aug 15th code freeze.
register. It enables Zoom Video. It appears that on at least one
card that Monzoon is using sets these bits by default. Nothing works
when these bits are set, everything works when they are clear.
Add commentary on some of the ti bits. Make code a little clearer.
Also remove a call to pcic_pci_pd6729 which was prematurely added in
the last commit.
linux_getdents uses VOP_READDIR( ..., &ncookies, &cookies ) instead of
VOP_READDIR( ..., NULL, NULL ) because it seems to need the offsets for
linux_dirent and sizeof(dirent) != sizeof(linux_dirent)...
PR: 29467
Submitted by: Michael Reifenberger <root@nihil.plaut.de>
Reviewed by: phk
boot CPU. This was the reason reboots on SMP systems could result in
weird hangs. Unlike the x86, we do not need to switch back to the boot
CPU in order to reboot the machine. See Section 3.4.5 of Part III
(Console Interface Architecture) from the Alpha Architecture Reference
Manual (aka the Brown Book) for more info.
common_attach is wrong as common attach initialize some fileds used by
mediainit routine. This was hard to notify because loading driver as kld
lead to mediainit routine being called after common_attach, though probe_phy
is called before.
MFC after: 1 week
traps, so that ddb can keep control (almost) no matter how it is
entered. This breaks time-critical interrupts while the system is
stopped in ddb, but I haven't noticed any significant problems except
that applications become confused about the time. Lost time will be
adjusted for later. Anyway, the half-baked disabling of interrupts in
Debugger() gives the same problems for the usual way of entering ddb.
bug for bug compatibility to ddb trap handlers after fixing the debugger
trap gates to be interrupt gates, but the fix was never committed. Now
I want the fix to apply to ddb.
- fix segment limit mis-calculation for GCODE_SEL, GDATA_SEL, GPRIV_SEL,
LUCODE_SEL and LUDATA_SEL.
- move `loader(8) metadata' related printf() after cninit().
- use atop macro (address to pages) for segment limit calculation
instead of i386_btop macro (bytes to pages).
- fix style bugs for the declarations of ints.
Reviewed by: bde, msmith (and arch & audit ML)
cdevsw entries have been for a long time.
Discover that we now have two version sof the same structure.
I will shoot one of them shortly when I figure out why someone thinks
they need it. (And I can prove they don't)
(netinet/ipprotosw.h should GO AWAY)
structure is always free()ed yet only sometimes malloc()ed. In particular,
it was simply set to point to l_filename from the a linker_file_t in
link_elf_link_preload_finish(). The l_filename had been malloc()ed inside
the kern_linker.c module and was being free()ed twice: once by
link_elf_unload_file() and again by linker_file_unload(), leading to
a panic.
How to duplicate the problem:
- Pre-load a kernel module from the loader, i.e. if_sis.ko
- Boot system
- Attempt to unload module with kldunload if_sis
- Bewm
The problem here is that the case where the module was loaded with kldload
after system boot would work correctly, so this bug went unnoticed until
I stubbed my toe on it just now. (Also, you can only trip this bug if
you compile a kernel with options DDB, but that's the default now.)
Fix: remember to malloc() a separate copy of the module name for the
l_name member of the gdb linkage structure in three places where the
linkage structure can be initialized.
flags with interrupts disabled to see if we should call ast() during
doreti. This was mostly submitted by Bruce, but his original patch did
the looping in ast() in assembly rather than in the ast() function itself.
Once we've actually called into the ast() function, it's cheaper to just
loop inside the function rather than returning from the function,
performing the check, and then calling the function again. However, we
can optimize the first check to avoid calling the function at all.
Other architectures may choose to implement this optimization if they
wish but it is not required for correct operation.
Submitted by: bde
the process of exiting the kernel. The ast() function now loops as long
as the PS_ASTPENDING or PS_NEEDRESCHED flags are set. It returns with
preemption disabled so that any further AST's that arrive via an
interrupt will be delayed until the low-level MD code returns to user
mode.
- Use u_int's to store the tick counts for profiling purposes so that we
do not need sched_lock just to read p_sticks. This also closes a
problem where the call to addupc_task() could screw up the arithmetic
due to non-atomic reads of p_sticks.
- Axe need_proftick(), aston(), astoff(), astpending(), need_resched(),
clear_resched(), and resched_wanted() in favor of direct bit operations
on p_sflag.
- Fix up locking with sched_lock some. In addupc_intr(), use sched_lock
to ensure pr_addr and pr_ticks are updated atomically with setting
PS_OWEUPC. In ast() we clear pr_ticks atomically with clearing
PS_OWEUPC. We also do not grab the lock just to test a flag.
- Simplify the handling of Giant in ast() slightly.
Reviewed by: bde (mostly)
