receiver was not sending an immediate ack with delayed acks turned on
when the input buffer is drained, preventing the transmitter from
restarting immediately.
Propogate the TCP_NODELAY option to accept()ed sockets. (Helps tbench and
is a good idea anyway).
Some cleanup. Identify additonal issues in comments.
MFC after: 1 day
released management apps.
1. Implement poll(). This will check for queued aif's so that a
subsequent ioctl call to retrieve the next aif will not block.
2. Don't catch signals when sleeping on a fib sent from userland. This
causes a race and panic due to the pthread context switcher waking
up the tsleep at inopportune times.
3. Fix some whitespace nits.
MFC after: 3 days
This API is supported in SVR4.0MP, Solaris, Linux, AIX and Tru64 Unix.
PR: bin/27489
Submitted by: Arun Sharma <arun@sharma-home.net>
Reviewed by: bde (prior version)
o Hide nonstandard functions and types in <netinet/in.h> when
_POSIX_SOURCE is defined.
o Add some missing types (required by POSIX.1-200x) to <netinet/in.h>.
o Restore vendor ID from Rev 1.1 in <netinet/in.h> and make use of new
__FBSDID() macro.
o Fix some miscellaneous issues in <arpa/inet.h>.
o Correct final argument for the inet_ntop() function (POSIX.1-200x).
o Get rid of the namespace pollution from <sys/types.h> in
<arpa/inet.h>.
Reviewed by: fenner
Partially submitted by: bde
you run out of mbuf address space.
kern/subr_mbuf.c: print a warning message when mb_alloc fails, again
rate-limited to at most once per second. This covers other
cases of mbuf allocation failures. Probably it also overlaps the
one handled in vm/vm_kern.c, so maybe the latter should go away.
This warning will let us gradually remove the printf that are scattered
across most network drivers to report mbuf allocation failures.
Those are potentially dangerous, in that they are not rate-limited and
can easily cause systems to panic.
Unless there is disagreement (which does not seem to be the case
judging from the discussion on -net so far), and because this is
sort of a safety bugfix, I plan to commit a similar change to STABLE
during the weekend (it affects kern/uipc_mbuf.c there).
Discussed-with: jlemon, silby and -net
the administrator to restrict access to the kernel message buffer.
It defaults to '1', which permits access, but if set to '0', requires
that the process making the sysctl() have appropriate privilege.
o Note that for this to be effective, access to this data via system
logs derived from /dev/klog must also be limited.
Obtained from: TrustedBSD Project
Sponsored by: DARPA, NAI Labs
that new models can inhabit kern.security.<modelname>.
o While I'm there, shorten somewhat excessive variable names, and clean
things up a little.
Obtained from: TrustedBSD Project
Sponsored by: DARPA, NAI Labs
- set sc->acpi_s4bios to 1 by default for hibernation until
OS-initiated S4 transition is implemented.
- change the behavior of acpi_sleep_state_sysctl() if new value is
the same as old one, do nothing instead of EINVAL.
interface address, blow the address away again before returning the
error.
In in_ifinit(), if we get an error from rtinit() and we've also got
a destination address, return the error rather than masking EEXISTS.
Failing to create a host route when configuring an interface should
be treated as an error.
pointless and would be inadequate for SMP systems. We will rely on the
VM system's locks to serialise this for now.
* Change pmap_remove() so that if the range being removed is larger than
the number of pages mapped by the pmap, we iterate over the currently
mapped pages instead of over the virtual address range. This should
make a difference when removing large virtual address ranges from an
address space.
mbuf allocation fails, and fix (i hope) a couple of style bugs.
I believe these printf() are extremely dangerous because now they can
occur on every incoming packet and are not rate limited. They were
meant to warn the sysadmin about lack of resources, but now they
can become a nice way to panic your system under load.
Other drivers (e.g. the fxp driver) have nothing like this.
There is a pending discussion on putting this kind of warnings
elsewhere, and I hope we can fix this soon.
underlying unaligned bcopy) on incoming packets that are already
available (albeit unaligned) in a buffer.
The performance improvement varies, depending on CPU and memory
speed, but can be quite large especially on slow CPUs. I have seen
over 50% increase on forwarding speed on the sis driver for the
486/133 (embedded systems), which does exactly the same thing.
The behaviour is controlled by a sysctl variable, hw.dc_quick which
defaults to 1. Set it to 0 to restore the old behaviour.
After running a few experiments (in userland, though) I am convinced
that doing the m_devget() is detrimental to performance in almost
all cases.
Even if your CPU has degraded performance with misaligned data,
the bcopy() in the driver has the same overhead due to misaligment
as the one that you save in the uiomove(), plus you do one extra
copy and pollute the cache.
But more often than not, you do not even have to touch the payload,
e.g. when you are forwarding packets, and even in the often-cited
case of NFS, you often end up passing a pointer to the payload to
the disk controller.
In any case, you can play with the sysctl variable to toggle between
the two behaviours, and see if it makes a difference.
MFC-after: 3 days
readability.
o Conditionalize only the SYSCTL definitions for the regression
tree, not the variables itself, decreasing the number of #ifdef
REGRESSIONs scattered in kern_mib.c, and making the code more
readable.
Sponsored by: DARPA, NAI Labs
have alignment problems.
On small boxes (e.g. the net4501 from Soekris, featuring a 486/133)
this provides huge performance benefits: the peak forwarding rate
with avg.sized packets goes up by 50-70% because of this change
alone. Faster CPUs might benefit less from this change, but in any
case the CPU has better things to do than waste time on useless
memory-to-memory copies.
Several drivers (for Tulip-like cards) might benefit from a similar
change.
Right now the new behaviour is controlled by a sysctl variable,
hw.sis_quick which defaults to 1 (on), you can set it to 0 to
reintroduce the old behaviour (and compare the results). The
variable is only there to show how much you can gain with this
change, it will go away soon.
Also, slightly simplify the code to initialize the ring buffers,
and remove a couple of dangerous printf's which could trigger on
any packet in case of mbuf shortage.
MFC-after: 3 days
received on an interface without an IP address, try to find a
non-loopback AF_INET address to use. If that fails, drop it.
Previously, we used the address at the top of the in_ifaddrhead list,
which didn't make much sense, and would cause a panic if there were no
AF_INET addresses configured on the system.
PR: 29337, 30524
Reviewed by: ru, jlemon
Obtained from: NetBSD
respects locks. Before SMPng, one was able to call psignal()
using the "call" command, but this is no longer possible because it
does not respect locks by itself. This is very useful when one has
gotten their machine into a state where it is impossible to spawn
ps/kill or su to root.
In this case, respecting locks essentially means trying to aquire the
proc lock before calling psignal(). We can't block in the debugger,
so if trylock fails, the operation fails. This also means that we
can't use pfind(), since that will attempt to lock the process for us.
Reviewed by: jhb
idle and the driver would not detect the event, requiring userland
to cycle the interface to bring it up again.
The fix consists in adding SIS_IMR_RX_IDLE to the interrupt mask and
add a command in sis_intr() to restart the receiver when this happens.
While at it, make the test of status bits more efficient.
for passive mode data connections (PASV/EPSV -> 227/229). Well,
the actual punching happens a bit later, when the aliasing link
becomes fully specified.
Prodded by: Danny Carroll <dannycarroll@hotmail.com>
MFC after: 1 week
The problem was that the ISO9660 code wasn't opening the device prior to
issuing ioctl calls. In particular, the device must be open before
iso_get_ssector() is called in iso_mountroot().
If the device isn't opened first, the disk layer blows up due to an
uninitialized variable.
The solution was to open the device, call iso_get_ssector() and then close
it again.
The ATAPI CDROM driver doesn't have this problem because it doesn't use the
disk layer, and evidently doesn't mind if someone issues an ioctl without
first issuing an open call.
Thanks to phk for pointing me at the source of this problem.
Tested by: dirk
MFC after: 1 week
It is legal to have a device with device type 0x1f, that just means
that the device is of unknown type. Instead, only check the peripheral
qualifier when deciding whether or not to reject a device based on its
inquiry information.
Tested by: julian
MFC after: 3 weeks
Merge in the irq 0 detection. Add comment about why.
If we have irq 0, ignore it like we do irq 255. Some BIOS writers aren't
careful like they should be.
allows us to move the sio softc data structure back into sio.c and
reduce the complexity of the non sio.c sio files.
Submitted by: bde
# I didn't fix the locking issues that bruce also submitted.
loader parameter. This allows us to more easily boot on big memory
configuration machines. hw.pccbb.start_mem. Reflect this in a sysctl
so we can read it from userland.
# Note: we need a TUNABLE_ULONG to do this right. I'll add that to
# kernel.h soon.
request to one that's supported by the bridge. I'm not 100% sure this
is correct, but it makes it easier for the cardbus bridge to allocate
its memory.
Similar code is needed for the I/O range. Also, I'm not sure if I
should be doing this based on memory or pmemory (but likely should do
it based on some flag that tells us to prefetch or not).
Talked about a long time ago with: msmith
NEWCARD. Other patches may be reqiured to sio to prevent a hang on
eject. Also add commented out entries for sio_pccard.c in files.pc98
to match other architectures.
Submitted by: yamamoto shigeru-san
changing. Also change it from 0x44000000 to 0x84000000 for large memory
machines.
# the PCI bus code should do this for us. This is a bandaide, not a
# solution.
click) do not include newline into the buffer. This is exacly how
things worked before my recent changes to the cut'n'paste code and
how they work in 4-STABLE.
pathconf() variables for directories, and set st_size and st_blocks
(of struct stat) for directories as appropriate. Note that st_size is
always set to DEV_BSIZE, since the size of the directories is not
currently kept.
Reviewed by: phk, bde
For an object type, we maintain a variable mb_mapfull. It is 0 by default
and is only raised to 1 in one place: when an mb_pop_cont() fails for
the first time, on the assumption that the reason for the failure is
due to the underlying map for the object (e.g. clust_map, mbuf_map) being
exhausted.
Problem and Changes:
Change how we define "mb_mapfull." It now means: "set to 1 when the first
mb_pop_cont() fails only in the kmem_malloc()-ing of the object, and
only if the call was with the M_TRYWAIT flag." This is a more conservative
definition and should avoid odd [but theoretically possible] situations
from occuring. i.e. we had set mb_mapfull to 1 thinking the map for the
object was actually exhausted when we _actually_ failed in malloc()ing
the space for the bucket structure managing the objects in the page
we're allocating.
card is ejected while we're in this routine.
yamamoto-san's original patch had a small race window for AX88190
chips, which I corrected by limiting the number of iterations we'd try
to reset the bits to be about 15ms rather than forever. This seems to
work for me, but I don't have a large collections of cards based on
this chipset.
Submitted by: YAMAMOTO Shigeru
snooped on. This causes all kinds of Bad Things(tm) to happen since
closing one session will clobber state that's needed for the other
one. This could theoretically be supported if the code was careful,
but until somebody implements that, preventing this will stop people
from unknowingly shooting themselves in the foot.
device cloned, and assign all further devices to depend on it. This
allows us to call dev_depends() on it at module unload time to get rid
of /dev/snp* (in the devfs case, anyway). For this to work, we must
not destroy the device at close time. [Idea stolen from if_tun.]
The above has the following sideaffects: (a) The snp device used by
watch(8) will remain after watch(8) exits. This is probably how it
should have been all along, and how it was before devfs came along.
(b) Module unload doesn't panic if there are any /dev/snp* devices
which haven't been used (and thus previously destroyed). Thus, we can
reenable the unload functionality disabled in rev. 1.65.
PR: 32012
MFC after:3 days
- Add memory barrier definition for sparc64.
Patch sent by David E. O'Brien, approved by maintainer.
- Fix an endianization error of a bus physical address used from SCRIPTS
that made the driver fail on big endian machines as sparc64.
The kernel certainly doesn't use _PATH_DEV or even /dev/ to find the device.
It cannot, since "/" has not been mounted. Maybe the only affect of using
/dev/ is that it gets put in the mounted-from name for "/", so that mount(8),
etc., display an absolute path before "/" has been remounted. Many have
never bothered typing the full path, and code that constructs a path in
rootdevnames[] never bothered to construct a full path, so the example
shouldn't have it.
Submitted by: bde
signature, but otherwise behaves just like a normal USB mass-storage
device. Add a new quirk to cover this case, and enable it for C-1
cameras. The quirk enables translation from the C-1 signature to
the normal CSWSIGNATURE value.
Reviewed by: n_hibma
to be followed by nfsnodehashtbl, so bzeroing callouts beyond the end of
tcp_syncache soon caused a null pointer panic when nfsnodehashtbl was
accessed.
Actually this porting supports Pegasus II chip so I guess some other
devices supported by NetBSD also work. But the devices list are not
included because I cannot confirm if they work.
Obtained from: NetBSD
MFC after: 3 weeks
Alpha can verify that it compiles fine there, too, this should be moved
to the MI section (or that problem fixed); I've only had x86 hardware to
est with.
Sponsored by: DARPA, NAI Labs (CBOSS project)
On OFW based machines, it is just too confusing having the firmware and
OS loader giving the same prompt. This is a nice compromise that 99% of the
users on non-OFW platforms will probably not even notice.
to the code for translating socket and private ioctls:
- Only perform socket ioctl translation if the file descriptor is a
socket.
- Treat socket ioctls on non-sockets specially, and for now assume
that these are directed at a tap/vmnet device, so translate the
ioctl numbers as appropriate (the way if_tap abuses some socket
ioctls to pass non-ifreq data is utterly bogus, but this is how
VMware on FreeBSD has always "worked"; I will deal with this
later).
- Add (untested) support for translating SIOCSIFADDR.
- In all cases where we fail to translate an ioctl, return ENOIOCTL
so that other handlers have a chance to do the translation.
This should fix the "/dev/vmnet1: Invalid argument" errors that
users of VMware were experiencing, though I have only verified this
on RELENG_4.
Submitted by: des (mostly)
MFC after: 3 days
2. Add OF_getprop_alloc(), a helper function that will malloc() a sufficient
amount of memory and then retrieve a property value into it.
Approved by: benno
Obtained from: NetBSD (1)
disabled unless verbose flag is set. Also fix some messages in terms
of English.
The critical messages and error messages in probe/attach routine are
unchanged by this commit.
but time and other interests is making it hard. Open the door for
new blood and fresh tactics now that the Linuxulator has had its
facelift.
Thanks to all who contributed during my tour of duty!
in asm files.
2. Temporarily cause subnormal operands in floating point operations
to be treated as zeros so that comlpetion of the operation does not
need to be emulated.
3. Catch fp_exception_other and correctly skip over the unfinished
instruction, but basically ignore them. Emulating the instruction
is not yet supported.
4. Zero td_retval[1] as well in syscall().
Submitted by: tmm (2, 3)
2. Add a TF_DONE macro, which fiddles a trapframe to make the retry on
return from traps act like a done (advance past the trapping
instruction instead of re-executing).
3. Flush the windows before entering the debugger, since it is no
longer done in the breakpoint trap vector.
4. Print a warning if trace <pid> is attempted, it is not yet implemented.
5. Print traps better and decode system calls in traces.
Submitted by: rwatson (4)
to determine if a process is using floating point. in order to avoid
sign extending a 13 bit immediate.
2. We don't need to context switch cwp anymore, it is better to just
fiddle the save tstate on return from traps. See exception.s 1.10
and 1.12.
3. Completely remove pcb_cwp.
4. Implement vmapbuf, vunmapbuf and vm_fault_quick. Completely remove
TODOs from vm_machdep.c (yay!).
Submitted by: tmm (1, 3, 4)
Obtained from: existing archs (4)
space from kernel space and from an alternate address space to kernel
space.
2. Remove the unused and unprototyped physcopy() and physzero() and replace
with the more versatile ascopy() and aszero(), inspired by the above.
These can be used to copy and zero physical pages of memory without mapping
them into kernel space first.
3. Use magic numbers for the offsets in the jmpbuf structure like other
platforms.
4. Use SET.
Submitted by: tmm (1, 4)
in the window trap vectors were mixed up. All this did is cause unnecesary
traps and look wierd in traces. Superfluous traps happen a lot in normal
operation, so we are rather good at recovering from them.
2. Store the arguments for a ktr trace in the right place.
3. Use a generic trap vector for breakpoints. It should not be special.
4. Save the frame pointer in the trap frame for kernel traps if DDB is compiled
in, otherwsie we don't save the out registers for kernel traps and stack
traces can't go through nested traps.
5. Apply the same fix to the return from kernel mode trap code as for user
mode traps. Ensure that the window we're returning to is the same one
that we restore to by fiddling the cwp in the saved tstate. This requires
that we transfer the values loaded from the trap frame into alternate
globals before restore-ing, but doing so is not very expensive and not
worth worrying about. Not changing the saved cwp can result in the register
values magically changing on return from traps if we happen to have slept
and the windows don't work out exactly the same. Fix the trace just before
the retry to account for different register usage.
6. Use a SET macro for loading address constants rather than a variation of
set and setx. set only works for 32 bit constants, while setx works for
64 bit constants as well, but produces bloated code when unnecessary.
Gas always generates the canonical 2 register, 6 instruction form, even
when it could be optimized; set uses 1 register and 2 instructions. At
the moment we assume that the kernel binary is below 4GB so set is
always sufficient, but the macro allows it to be configured. Note that
this has nothing to do with 32 vs. 64 bit address space, it only applies
to addresses of symbols which are known at compile/link time.
Submitted by: tmm (6)
time in the cases where it really sends the drive out to lunch, but it also
allows us to catch very wierd edge cases of strange drives that might take
a very long time (emulated disk drives over a network, e.g.).
vnodes. This will hopefully serve as a base from which we can
expand the MP code. We currently do not attempt to obtain any
mutex or SX locks, but the door is open to add them when we nail
down exactly how that part of it is going to work.
alpha pmap. In particular -
- pd_entry_t and pt_entry_t are now u_int32_t instead of a pointer.
This is to enable cleaner PAE and x86-64 support down the track sor
that we can change the pd_entry_t/pt_entry_t types to 64 bit entities.
- Terminate "unsigned *ptep, pte" with extreme prejudice and use the
correct pt_entry_t/pd_entry_t types.
- Various other cosmetic changes to match cleanups elsewhere.
- This eliminates a boatload of casts.
- use VM_MAXUSER_ADDRESS in place of UPT_MIN_ADDRESS in a couple of places
where we're testing user address space limits. Assuming the page tables
start directly after the end of user space is not a safe assumption.
There is still more to go.
a page boundary, since we've already allocated all our contiguous kva
space up front. This eliminates some memory wastage, and allows us to
actually reach the # of objects were specified in the zinit() call.
Reviewed by: peter, dillon
Protect against an infinite loop when prefaulting pages. This can
happen when the vm system maps past the end of an object or tries
to map a zero length object, the pmap layer misses the fact that
offsets wrap into negative numbers and we get stuck.
Make flushing dirty pages work correctly on filesystems that
unexpectedly do not complete writes even with sync I/O requests.
This should help the behavior of mmaped files when using
softupdates (and perhaps in other circumstances also.)
- use NPTEPG/NPDEPG instead of magic 1024 (important for PAE)
- use pt_entry_t instead of unsigned (important for PAE)
- use vm_offset_t instead of unsigned for va's (important for x86-64)
ill effects. This should fix problems threaded programs are having with
auto-detecting CPU type.
Reported by: Joe Clarke <marcus@marcuscom.com>
Tested by: Joe Clarke <marcus@marcuscom.com>
Reviewed by: jhb
MFC after: 1 week
src/sys/dev/an/if_an_pccard.c rev 1.10 change requires to include
"card_if.h" but Makefile doesn't know about this file. Without this,
kernel build will fail at this driver.
lock as this usually makes the problem worse.
- If we get a page fault while holding a spin lock, treat it as a fatal
trap and don't even bother calling into the VM since calling into the
VM will panic when trying to lock Giant before we can get a useful
message anyways.
Now you can say;
# sysctl hw.acpi.lid_switch_state=NONE
instead of specifying unsupported _Sx object in the system.
Actually, S4B is going to disappear in ACPICA and we already have
hw.acpi.s4bios to distinguish BIOS hibernation or OS hibernation.
pccard layer and rename them PCMCIA_CARD and PCMCIA_CARD2 respectively
(note, this is being done with an eye towards NetBSD integreation so
it is easier to keep lists of cards between us and them in sync).
Use this in the an and wi drivers.
reference: with td->td_ucred, it will be desirable to authorize
based on td->td_ucred, rather than p->p_ucred.
o Since the same variable 'p' was later used with pfind() on the target
process for the wakeup, introduce a new local variable 'targetp'
to use instead.
Obtained from: TrustedBSD Project
Sponsored by: DARPA, NAI Labs
compare the slot offset against the predicted offset, not a boolean
flag. This typo effectively disabled the sequential optimisation,
but was otherwise harmless.
Not surprisingly, fixing this improves performance in the sequential
access case. I am seeing a 7% speedup on one machine here; using
dirhash when sequentially looking up directory entries is now about
5% faster instead of 2% slower than the non-dirhash case.
Submitted by: KOIE Hidetaka <koie@suri.co.jp>
MFC after: 1 week
Replace uses of holdfp() with fget*() or fgetvp*() calls as appropriate
introduce fget(), fget_read(), fget_write() - these functions will take
a thread and file descriptor and return a file pointer with its ref
count bumped.
introduce fgetvp(), fgetvp_read(), fgetvp_write() - these functions will
take a thread and file descriptor and return a vref()'d vnode.
*_read() requires that the file pointer be FREAD, *_write that it be
FWRITE.
This continues the cleanup of struct filedesc and struct file access
routines which, when are all through with it, will allow us to then
make the API calls MP safe and be able to move Giant down into the fo_*
functions.
- Restore inferior() to being iterative rather than recursive.
- Assert that the proctree_lock is held in inferior() and change the one
caller to get a shared lock of it. This also ensures that we hold the
lock after performing the check so the check can't be made invalid out
from under us after the check but before we act on it.
Requested by: bde
reading cr.ivr, as well as writing to cr.eoi.
o use global variables to pass information to os_boot_rendez
so that it doesn't have to jump through hoops to find it
out. This avoids traps on the AP without it even being
initialized. This fixes SMP configurations.
o Move the probing of the MADT to the end of cpu_startup,
instead of at the start of cpu_mp_probe. We need to probe
the MADT for non-SMP configurations as well. This fixes
uniprocessor configurations.
o Serialize AP wake-up by waiting for the AP. We need to do
this since we use global variables to for the AP to use.
As a side-effect, we can use printf() more easily to see
what's going on.
to avoid the need for rpc.lockd to perform client locks. Using
this option a user can revert back to using local locks for NFS mounts
like we did before we had rpc.lockd.
one to perform a vn_open using temporary/other/fake credentials.
Modify the nfs client side locking code to use vn_open_cred() passing
proc0's ucred instead of the old way which was to temporary raise
privs while running vn_open(). This should close the race hopefully.
trying something new with the macros here and will likely try to get
them adopted by NetBSD as well as moving them to other drivers. They
make the list more compact and easier to read, the price of rigid to a
schema for generating them (of course there are those that would argue
this isn't bad).
Obtained from: NetBSD's if_wi_pcmcia.c 1.9
off chip that was on one prototype board. However, this appears to be
a design that many chipsets are compatible with its PPEC register set
(eg the Omega 82c094). Through the kindness of the Red Hat developer
David Woodhouse, I now have this datasheet.
I may take the advise of one of the bsd-nomads (whose name
unfortunately escapes me at the moment) and split out all these 16-bit
I/O mapped PCI devices into a separate driver...
that it has one BAR that's mapped to 0x3e0 and is I/O only. It does
not conform to the Yenta spec, like other PCI PCMICA bridges do (eg
the TI 1031, which is mostly a 1131 w/o 32bit card support). It
appears that this chip may also need to not route PCI interrupts
as well.
This chip is used in the NEC Versa 2430CD (and it appears that
sometimes it works, while other times it doesn't) and others in the
2400 series. While the NEC website claims Cardbus support, I can't figure
out how that is possible.
Submitted by: Ben Timby <ben@webexc.com>
