This is an odd one. This patch appears to fix a panic related to background
bitmap writes (for FFS), though neither Kirk, Ian, or I can figure out how
B_CLUSTEROK could possibly be set on a bitmap block to cause the clustering
code to improperly cluster with a buffer undergoing a background write.
In anycase, the clustering code is very fragile and this patch helps with
that, as well as possibly fixing a bug Andre was having.
Suggested by: Ian Dowse <iedowse@maths.tcd.ie>
Testing by: Andre Albsmeier <andre.albsmeier@mchp.siemens.de>
All calls to mtx_init() for mutexes that recurse must now include
the MTX_RECURSE bit in the flag argument variable. This change is in
preparation for an upcoming (further) mutex API cleanup.
The witness code will call panic() if a lock is found to recurse but
the MTX_RECURSE bit was not set during the lock's initialization.
The old MTX_RECURSE "state" bit (in mtx_lock) has been renamed to
MTX_RECURSED, which is more appropriate given its meaning.
The following locks have been made "recursive," thus far:
eventhandler, Giant, callout, sched_lock, possibly some others declared
in the architecture-specific code, all of the network card driver locks
in pci/, as well as some other locks in dev/ stuff that I've found to
be recursive.
Reviewed by: jhb
attempt to read memory when siz is 0
- Clarify comments referring to strlcat() usage
PR: 24278, 24295
Submitted by: Tony Finch <dot@dotat.at>
Richard Kettlewell <rjk@greenend.org.uk>
Reviewed by: -audit
active memory maps. This removes the need to change the memory
map from common to attribute every time a packet is sent/received.
This increases performance and decreases cpu load (ping times on
slow machines improve by about 1.5ms).
Move out the old common memory/attrbiute memory hack functions to a
new header file to tidy up the main code. I want to keep them available
for a while.
non-386 atomic_load_acq(). %eax is an input since its value is used in
the cmpxchg instruction, but we don't care what value it is, so setting
it to a specific value is just wasteful. Thus, it is being used without
being initialized as the warning stated, but it is ok for it to be used
because its value isn't important. Thus, we are only sort of lying when
we say it is an output only operand.
- Add "cc" to the clobber list for atomic_load_acq() since the cmpxchgl
changes ZF.
slow enough as it is, without having to constantly check that it really
is an i386 still. It was possible to compile out the conditionals for
faster cpus by leaving out 'I386_CPU', but it was not possible to
unconditionally compile for the i386. You got the runtime checking whether
you wanted it or not. This makes I386_CPU mutually exclusive with the
other cpu types, and tidies things up a little in the process.
Reviewed by: alfred, markm, phk, benno, jlemon, jhb, jake, grog, msmith,
jasone, dcs, des (and a bunch more people who encouraged it)
compiling errors where gcc would run out of registers.
- Add "cc" to the list of clobbers for micro-ops where we perform
instructions that alter %eflags.
- Use xchgl instead of cmpxchgl to release a spin lock. This could allow
for more efficient register allocation as we no longer mandate that %eax
be used.
- Reenable the optimized mutex micro-ops in the non-i386 case.
allocation routines are being called safely. Since we drop our relevant
mbuf mutex and acquire Giant before we call kmem_malloc(), we have
to make sure that this does not pave the way for a fatal lock order
reversal. Check that either Giant is already held (in which case it's safe
to grab it again and recurse on it) or, if Giant is not held, that no
other locks are held before we try to acquire Giant.
Similarily, add a KASSERT valid in the WITNESS case in m_reclaim() to
nail callers who end up in m_reclaim() and hold a lock.
Pointed out by: jhb
we *really* are.
It should be noted that there is a degenerate case where soft tape
location will be lost (not causing a frozen state- but causing
the loss of reporting fileno/blockno)- that's where you backspace
over a filemark- you stop backspacing as soon as you cross the
filemark, but you have no idea what the record number now is because
you have no idea how many records you are into the file you just
backed into. Such is life.
While I'm at it, also pick up residuals from writing filemarks.
PR: 24222
that modules can call.
- Remove the old gcc <= 2.8 versions of the atomic ops.
- Resort the order of some things in the file so that there is only
one #ifdef for KLD_MODULE, and so that all WANT_FUNCTIONS stuff is
moved to the bottom of the file.
- Remove ATOMIC_ACQ_REL() and just use explicit macros instead.
only CCB type but also extra flags- one of which can be "position
updated".
In other changes: Add in a SA_QUIRK_NO_CPAGE quirk so that it's possible
to avoid using a (broken) device's implementation of he DEVICE COMPRESSION
page.
Also do a couple of printout cleanups.
As per some discussion on FreeBSD-scsi, skip doing tape flushing
if we're reading tape logical block location (MTIOCRDSPOS).
(so we can see rapidly whether something was a fabric device but is
now gone).
Add a tag which says what role this adapter should take. It can take
on the value of None, Target, Initiator or Both. None is useful for
warm failover purposes. Remove the ISP_CFG_NOINIT silliness since
a role of "None" does this.
Add a isp_lastmbxcmd tag to store the opcode for the last mailbox
command used.
Module) and FBM (Fibre Buffer Modules). Also remember to clear the
semaphore registers. Tell the RISC processor to not halt on FPM
parity errors.
Throw out the ISP_CFG_NOINIT silliness and instead go to the use of
adapter 'roles' to see whether one completes initialization or not
(mostly for Fibre Channel). The ultimate intent, btw, of all of this
is to have a warm standby adapter for failover reasons. Because
we do roles now, setting of Target Capable Class 3 service parameters
in the ICB for the 2x00 cards reflects from role. Also, in isp_start,
if we're not supporting an initiator role, we bounce outgoing commands
with a Selection Timeout error. Also clean out the TOGGLE_TMODE
goop for FC- there is no toggling of target mode like there is
for parallel SCSI cards.
Do more cleanup with respect to using target ids 0..125 in F-port
topologies. Also keep track of things which *were* fabric devices
so that when you rescan the fabric you can notify the outer layers
when fabric devices go away.
Only force a LOGOUT for fabric devices if they're still logged in
(i.e., you cat their Port Database entry. Clean up the Get All Next
scanning.
Finally, use a new tag in the softc to store the opcode for the
last mailbox command used so we can report which opcode timed
out.
in-core pointers to summary information. An array in this region
(fs_csp) could overflow on filesystems with a very large number of
cylinder groups (~16000 on i386 with 8k blocks). When this happens,
other fields in the superblock get corrupted, and fsck refuses to
check the filesystem.
Solve this problem by replacing the fs_csp array in 'struct fs'
with a single pointer, and add padding to keep the length of the
128-byte region fixed. Update the kernel and userland utilities
to use just this single pointer.
With this change, the kernel no longer makes use of the superblock
fields 'fs_csshift' and 'fs_csmask'. Add a comment to newfs/mkfs.c
to indicate that these fields must be calculated for compatibility
with older kernels.
Reviewed by: mckusick
broken the handling of uncompressed VJ packets. The attached diff should
hopefully fix that.
Submitted by: Thomas Moestl <tmoestl@gmx.net>
Reviewed by: Sergio de Souza Prallon <prallon@tmp.com.br>
time I tinkered around here. Since INTREN is called from the interrupt
critical path now, it should not be too expensive. In this case, we
look at the bits being changed to decide which 8 bit IO port to write to
rather than unconditionally writing to both. I could probably have gone
further and only done the write if the bits actually changed, but that
seemed overkill for the usual case in interrupt threads.
[an outb is rather expensive when it has to cross the ISA bus]
exactly the same functionality via a sysctl, making this feature
a run-time option.
The default is 1(ON), which means that /dev/random device will
NOT block at startup.
setting kern.random.sys.seeded to 0(OFF) will cause /dev/random
to block until the next reseed, at which stage the sysctl
will be changed back to 1(ON).
While I'm here, clean up the sysctls, and make them dynamic.
Reviewed by: des
Tested on Alpha by: obrien
implement memory fences for the 486+. The 386 still uses versions w/o
memory fences as all operations on the 386 are not program ordered.
The 386 versions are not MP safe.
not yet been caught), don't save the config with a null drive
name (which causes the drive to be renamed "plex" on the next
start), put in the text "*invalid*" instead.
This is damage control, not a fix.
Experienced by: peter
Break some long format strings so that they fit in style(9)-sized
lines.
Remove some "outdentation".
while we are copying it to the kinfo_proc structure.
- Test against p_stat to see if we are blocked on a mutex.
- Terminate ki_mtxname with a null char rather than ki_wmesg.
declarations of a variable of the same name. The one in the outer block
was unused and probably just slipped in at one point or another. This
silences a compiler warning.
#include "quad.h"
to:
#include <libkern/quad.h>
as the former breaks under a kernel build.
This change had already been performed on the files that were used in other
kernel builds, however the PowerPC kernel build seems to require some that
weren't being used.
Reviewed by: obrien, peter
all of the hwvol members of struct snd_mixer live in a hwvol_* namespace.
- When changing the mixer device via the hwvol_mixer sysctl, reset the
muted state so that a mute operation on a new device won't try to
unmute the new device with the old device's saved volume.
- When the volume is muted, if a down or up volume request is received,
first restore the saved volume level and then adjust it.
Reviewed by: cg
from a node, but does it via the locking queue, thus ensuring that the
node is locked when it's hook is removed.
Add 'deadnode' and 'deadhook' structures for when a node or hook is
invalidated but not yet freed. (not yet freed)
__FreeBSD_version 500015 can be used to detect their disappearance.
- Move the symbols for SMP_prvspace and lapic from globals.s to
locore.s.
- Remove globals.s with extreme prejudice.
from Traverse Technology and also the Teles PCI-TJ cards both based on the
chipset combination of the Siemens ISAC and the TJNet Tiger300/320 chips.
The itjc/i4b_hdlc.h file will hopefully soon be merged with the file
/usr/src/sys/i4b/layer1/i4b_hdlc.h.
Submitted by: Sergio de Souza Prallon <prallon@tmp.com.br>
enables the driver to work on current (I have not tested the isa part fully,
but I suppose it should work). The patching from jlemon caused a crash at
probe time. This is probably my fault not having added a comment for
"RESET_SOFT" saying: "this resets all!".
Submitted by: Hans Petter Selasky <hselasky@c2i.net>
cause the watchdog to (incorrectly) detect the interface state
going from down to up one second after lmc_ifup() was called,
causing lmc_ifup() to be called again, and resetting everything.
Fix by checking the interface state within lmc_ifup().
Submitted by: Darren Croke <djc@packetdesign.com>
Rewrite lockrange and unlockrange. The lock table is now a fixed
size, so there is no possibility for race conditions when expanding.
The current size (256 locked ranges) should be large enough that it
makes no sense to expand it. To do expansion right would require
quiescing the plex (requiring at least 256 I/O completions), and the
performance implications are horrendous.
Add a mutex per plex for accessing the lock table.
Based on analysis by: tegge
This should eliminate one case of foot shooting .
vinum_scandisk: If a drive in the partition table is downed, free it.
This duplicates code for the compatibility partition, which for some
reason was omitted here.
m_reclaim() and re-acquire it when m_reclaim() returns. This means that
we now call the drain routines without holding the mutex lock and
recursing into it. This was done for mainly two reasons:
(i) Avoid the long recursion; long recursions are typically bad and this
is the case here because we block all other code from freeing mbufs
if they need to. Doing that is kind of counter-productive, since we're
really hoping that someone will free.
(ii) More importantly, avoid a potential lock order reversal. Right now,
not all the locks have been added to our networking code; but
without this change, we're introducing the possibility for deadlock.
Consider for example ip_drain(). We will likely eventually introduce
a lock for ipq there, and so ip_freef() will be called with ipq lock
held. But, ip_freef() calls m_freem() which in turn acquires the
mmbfree lock. Since we were previously calling ip_drain() with mmbfree
held, our lock order would be: mmbfree->ipq->mmbfree. Some other code
may very well lock ipq first and then call ip_freef(). This would
result in the regular lock order, ipq->mmbfree. Clearly, we have
deadlock if one thread acquires the ipq lock and sits waiting for
mmbfree while another thread calling m_reclaim() acquires mmbfree
and sits waiting for the ipq lock.
Also, make sure to add a comment above m_reclaim()'s definition briefly
explaining this. Also document this above the call to m_reclaim() in
m_mballoc_wait().
Suggested and reviewed by: alfred
one memory map. The memory window for the PCIC is identifed by the resource id
for NEWSBUS drivers. pccardd always uses window 0 and rid 0 when setting maps
up. This fix does not affect pccardd's handling of common memory for ed cards.
Reviewed by: imp
be 64 bits wide. The largest known current actual physical implementation
is 40 bits, so BUS_SPACE_MAXADDR should reflect this. It also seems to
me that BUS_SPACE_UNRESTRICTED should b ~0UL, not ~0.
avma1pp_attach_avma1pp, since the former may be called multiple
times and we only want to initialize the mutex once.
Submitted by: Thomas Moestl <tmoestl@gmx.net>
reserved and now allocated TCP flags in incoming packets. This patch
stops overloading those bits in the IP firewall rules, and moves
colliding flags to a seperate field, ipflg. The IPFW userland
management tool, ipfw(8), is updated to reflect this change. New TCP
flags related to ECN are now included in tcp.h for reference, although
we don't currently implement TCP+ECN.
o To use this fix without completely rebuilding, it is sufficient to copy
ip_fw.h and tcp.h into your appropriate include directory, then rebuild
the ipfw kernel module, and ipfw tool, and install both. Note that a
mismatch between module and userland tool will result in incorrect
installation of firewall rules that may have unexpected effects. This
is an MFC candidate, following shakedown. This bug does not appear
to affect ipfilter.
Reviewed by: security-officer, billf
Reported by: Aragon Gouveia <aragon@phat.za.net>
that require us to register our FC4 types of interest. Allow ourselves, in
F-port topologies, to start logging in fabric devices in the target 0..125
range. Change ISPASYNC_PDB_CHANGED (misnamed) to ISPASYNC_LOGGED_INOUT.
Fix (*SMACK*) again some default WWN stuff. This is *really* hard to get
right across all the range of platforms.
hscbs may be traded during error recovery due to the way
we manage the qinfifo. This has the effect of changing the
index to the sense buffer even though the request sense command
references the original buffer. SCBs don't play this swapping
game and so serve as a more consistent reference.
the pipe, then we were corrupting the pipe_zone free list by calling
pipeclose on rpipe twice. NULL out rpipe to avoid this.
Reviewed by: dillon
Reviewed by: iedowse
- add a dma hack similar to the NetBSD one
- change PDQ_OS_MEM{RD,WR} to use readl/writel rather than deref'ing
a 32 bit va.
Note that I did just enough to get this working on alpha. I probably
should have updated it to use busspace, but I was too lazy to navigate
the twisty minefield of ifdefs that make up this driver.
Tested by: wilko (on both x86 and alpha)
Add the entry for the Yano U640MO-03 MO drive. (ifdef-0-ed out for now)
Fix a hack were an original buffer was modified instead of copied
(cmd[] -> (*rcmd)[])
Submitted by: Iwasa Kazmi <kzmi@ca2.so-net.ne.jp>
instead of the requested length. Otherwise all transfers look like 0 byte
transfers to CAM.
Submitted by: Iwasa Kazmi <kzmi@ca2.so-net.ne.jp>
2) Finalise the command in the case of CBI transfers with CCI (command
completion interrupt).
3) Remove a redundant bzero of a buffer.
idea either) in ufs_extattr_rm.
o More completely fill out the local_aio structure when writing out the
zero'd extended attribute in ufs_extattr_rm -- previoulsy, this worked
fine, but probably should not have. This corrects extraneous warnings
about inconsistent inodes following file deletion.
Reviewed by: jedgar
ufs_extattr_rm.
o Make both reporting locations report the function name where the
inconsistency is discovered, as well as the inode number in question.
Reviewed by: jedgar
The pccard_function_init() call creates a bunch of inactive resources
that are persistant and configured on demand. When the child driver
"allocates" a resource it is connected up to one of these. When the
child releases the resource, we should not delete our copy, just
deactivate it again. Otherwise there is nothing to recreate it again
after several probe functions have run and done an alloc/release cycle.
INVARIANTS shows 0xdeadc0de without this.
More work is needed to do a sweep though the pccard_function_disable()
call to actually delete the resources for real. Right now, we leak
memory on eject (at best), so Dont Do That(TM) yet. This affects
16 bit pccards on a cardbus bridge only. This will be fixed soon, but
for now it gets the cards working.
Reviewed by: imp
to match the pccard.conf file. There are more ID's that need adding, but
these seem to be the common ones.
This was committed on an ep0 interface under NEWCARD:
ep0: <3Com 3c589 10Mbps Ethernet> at port 0x300-0x30f irq 9 function 0 config 1 on pccard1
ep0: Ethernet address 00:10:4b:df:48:57
Reviewed by: imp
attribute read--the offset is required to be 0 by an earlier check,
meaning that it will always be within the scope of the attribute data.
This change should have no impact on executed code paths other than
removing the unnecessary check: please report if any new failures
start to occur as a result.
Obtained from: TrustedBSD Project
to conform to style(9), plus one other convention that I use:
o Declare variables at the start of the function, rather than in blocks
when it doesn't help understanding (mine).
o 80 column limit.
o BSD style statement continuation, rather than "gnu" style.
bus to use. We need to set it here.
This fixes the problem where a probe routine establishes and
disestablishes the interrupt and then we get a panic in the probe
routine.
However, we pass the pointer to the interrupt hanlder count to the
parent bus, which writes its own cookie there, so there may be some
problems with that which isn't apparent at the moment.
Commit made from: laptop running NEWCARD with sn driver (which works,
but gets the wrong ethernet address).
o Fix OLDCARD to use the new interface.
o Rename the offsetp argument to deltap to more closely reflect what it
is returning (it returns the delta from the requested value to the actual
value).
o Remove duplicate $FreeBSD$ in pccbb.c
o Allow deltap to be NULL.
o Convert new isa pcic driver and add XXX comments that this function isn't
actually implemented there (which means that NEWCARD pccard stuff won't
work there until it is).
o Revert attempts to make old inferface work in NEWCARD.
Subitted by: peter (Parts of the new version code)
compiled under newcard yet. ep works just fine under newcard with the
missing ID matching code added (not committed yet):
ep0: <3Com 3c589 10Mbps Ethernet> at port 0x300-0x30f irq 9 function 0
config 1 on pccard1
ep0: Ethernet address 00:10:4b:df:48:57
