in practice, the error (currently) makes no difference because the computation
performed by KVADDR() hides the error. This revision fixes the error.
Also, eliminate a (now) unused definition.
maximum size of the kmem map can be greater than 4GB, there is little point
in making the kernel virtual address space larger than 6GB.
Tested by: kris@
Now that the pseudo-interface cloner has an internal list of instances,
there is no need to create a softc. The softc only contains a pointer to
the ifp, which means there is no valid reason to keep it. While there,
remove the corresponding malloc-pool.
Approved by: philip (mentor)
Adaptec RAID 2045
Adaptec RAID 2405
Adaptec RAID 2445
Adaptec RAID 2805
Without this change these devices are supported by the driver's family
support, but they then appear as "Adaptec RAID Controller" in boot
messages and the dev.aac.0.%desc sysctl.
This includes hotkeys support and sysctl variables to control camera
and card reader. These new sysctls don't have CTFLAG_ANYBODY set.
While there add entries to devd.conf related to the Eee volume keys.
Reviewed by: phillip
MFC after: 1 week
Also tested by: lme (previous version)
semaphores. Specifically, semaphores are now represented as new file
descriptor type that is set to close on exec. This removes the need for
all of the manual process reference counting (and fork, exec, and exit
event handlers) as the normal file descriptor operations handle all of
that for us nicely. It is also suggested as one possible implementation
in the spec and at least one other OS (OS X) uses this approach.
Some bugs that were fixed as a result include:
- References to a named semaphore whose name is removed still work after
the sem_unlink() operation. Prior to this patch, if a semaphore's name
was removed, valid handles from sem_open() would get EINVAL errors from
sem_getvalue(), sem_post(), etc. This fixes that.
- Unnamed semaphores created with sem_init() were not cleaned up when a
process exited or exec'd. They were only cleaned up if the process
did an explicit sem_destroy(). This could result in a leak of semaphore
objects that could never be cleaned up.
- On the other hand, if another process guessed the id (kernel pointer to
'struct ksem' of an unnamed semaphore (created via sem_init)) and had
write access to the semaphore based on UID/GID checks, then that other
process could manipulate the semaphore via sem_destroy(), sem_post(),
sem_wait(), etc.
- As part of the permission check (UID/GID), the umask of the proces
creating the semaphore was not honored. Thus if your umask denied group
read/write access but the explicit mode in the sem_init() call allowed
it, the semaphore would be readable/writable by other users in the
same group, for example. This includes access via the previous bug.
- If the module refused to unload because there were active semaphores,
then it might have deregistered one or more of the semaphore system
calls before it noticed that there was a problem. I'm not sure if
this actually happened as the order that modules are discovered by the
kernel linker depends on how the actual .ko file is linked. One can
make the order deterministic by using a single module with a mod_event
handler that explicitly registers syscalls (and deregisters during
unload after any checks). This also fixes a race where even if the
sem_module unloaded first it would have destroyed locks that the
syscalls might be trying to access if they are still executing when
they are unloaded.
XXX: By the way, deregistering system calls doesn't do any blocking
to drain any threads from the calls.
- Some minor fixes to errno values on error. For example, sem_init()
isn't documented to return ENFILE or EMFILE if we run out of semaphores
the way that sem_open() can. Instead, it should return ENOSPC in that
case.
Other changes:
- Kernel semaphores now use a hash table to manage the namespace of
named semaphores nearly in a similar fashion to the POSIX shared memory
object file descriptors. Kernel semaphores can now also have names
longer than 14 chars (up to MAXPATHLEN) and can include subdirectories
in their pathname.
- The UID/GID permission checks for access to a named semaphore are now
done via vaccess() rather than a home-rolled set of checks.
- Now that kernel semaphores have an associated file object, the various
MAC checks for POSIX semaphores accept both a file credential and an
active credential. There is also a new posixsem_check_stat() since it
is possible to fstat() a semaphore file descriptor.
- A small set of regression tests (using the ksem API directly) is present
in src/tools/regression/posixsem.
Reported by: kris (1)
Tested by: kris
Reviewed by: rwatson (lightly)
MFC after: 1 month
unsynchronized. While races were extremely rare, we've now had a
couple of reports of panics in environments involving large numbers of
IPSEC tunnels being added very quickly on an active system.
- Add accessor functions ifnet_byindex(), ifaddr_byindex(),
ifdev_byindex() to replace existing accessor macros. These functions
now acquire the ifnet lock before derefencing the table.
- Add IFNET_WLOCK_ASSERT().
- Add static accessor functions ifnet_setbyindex(), ifdev_setbyindex(),
which set values in the table either asserting of acquiring the ifnet
lock.
- Use accessor functions throughout if.c to modify and read
ifindex_table.
- Rework ifnet attach/detach to lock around ifindex_table modification.
Note that these changes simply close races around use of ifindex_table,
and make no attempt to solve the probem of disappearing ifnets. Further
refinement of this work, including with respect to ifindex_table
resizing, is still required.
In a future change, the ifnet lock should be converted from a mutex to an
rwlock in order to reduce contention.
Reviewed and tested by: brooks
- Each log entry contains a text description in the "description" field of
the entry. The existing decode logic always ended up duplicating
information that was already in the description string. This made the
logs overly verbose. Now we just print out the description string.
- Add some simple parsing of the timestamp and event classes.
Reviewed by: ambrisko, scottl
MFC after: 2 weeks
- Fetch events from the controller in batches of 15 rather than a single
event at a time.
- When fetching events from the controller, honor the event class and
locale settings (via hw.mfi tunables). This also allows the firmware to
skip over unwanted log entries resulting in fewer requests to the
controller if there many unwanted log entries since the last clean
shutdown.
- Don't drop the driver mutex while decoding an event.
- If we get an error other than MFI_STAT_NOT_FOUND (basically EOF for
hitting the end of the event log) then emit a warning and bail on
processing further log entries.
Reviewed by: ambrisko, scottl
MFC after: 2 weeks
to INT_MAX. Otherwise, a process could create a semaphore (or increase
its value via ksem_post()) beyond INT_MAX and sem_getvalue() would return
a negative value. sem_getvalue() is only supposed to return a negative
value if that is the number of waiters for that semaphore.
MFC after: 2 weeks
provides the correct semantics for flock(2) style locks which are used by the
lockf(1) command line tool and the pidfile(3) library. It also implements
recovery from server restarts and ensures that dirty cache blocks are written
to the server before obtaining locks (allowing multiple clients to use file
locking to safely share data).
Sponsored by: Isilon Systems
PR: 94256
MFC after: 2 weeks
Bonus: including kern.mk just to pick kernel warning flags
was an extremely bad idea anyway, because it also picked
up CFLAGS (it probably wasn't the case at the time of CVS
rev. 1.1, I haven't checked). Remove duplicate CWARNFLAGS
from CFLAGS.
so we cannot compile it with -fstack-protector[-all] flags (or
it will self-recurse); this is ensured in sys/conf/files. This
OTOH means that checking for defines __SSP__ and __SSP_ALL__ to
determine if we should be compiling the support is impossible
(which it was trying, resulting in an empty object file). Fix
this by always compiling the symbols in this files. It's good
because it allows us to always have SSP support, and then compile
with SSP selectively.
Repoted by: tinderbox
- It is opt-out for now so as to give it maximum testing, but it may be
turned opt-in for stable branches depending on the consensus. You
can turn it off with WITHOUT_SSP.
- WITHOUT_SSP was previously used to disable the build of GNU libssp.
It is harmless to steal the knob as SSP symbols have been provided
by libc for a long time, GNU libssp should not have been much used.
- SSP is disabled in a few corners such as system bootstrap programs
(sys/boot), process bootstrap code (rtld, csu) and SSP symbols themselves.
- It should be safe to use -fstack-protector-all to build world, however
libc will be automatically downgraded to -fstack-protector because it
breaks rtld otherwise.
- This option is unavailable on ia64.
Enable GCC stack protection (aka Propolice) for kernel:
- It is opt-out for now so as to give it maximum testing.
- Do not compile your kernel with -fstack-protector-all, it won't work.
Submitted by: Jeremie Le Hen <jeremie@le-hen.org>
that modify condition codes (the carry bit, in this case). Without
"__volatile", the compiler might add the inline assembler instructions
between unrelated code which also uses condition codes, modifying the
latter.
This prevents the TCP pseudo header checksum calculation done in
tcp_output() from having effects on other conditions when compiled
with GCC 4.2.1 at "-O2" and "options INET6" left out. [1]
Reported & tested by: Boris Kochergin [1]
MFC after: 3 days
Now that st_rdev is being automatically generated by the kernel, there
is no need to define static major/minor numbers for the iodev and
memdev. We still need the minor numbers for the memdev, however, to
distinguish between /dev/mem and /dev/kmem.
Approved by: philip (mentor)
in_ifaddrhashtbl in in_ifinit because error handler in in_control removes
entries only for AF_INET addresses. If in_ifinit is called for the cloned
inteface that has just been created its address family is not AF_INET and
therefor LIST_REMOVE is not called for respective LIST_INSERT_HEAD and
freed entries remain in in_ifaddrhashtbl and lead to memory corruption.
PR: kern/124384
locked and unlocked completely in userland. by locking and unlocking mutex
in userland, it reduces the total time a mutex is locked by a thread,
in some application code, a mutex only protects a small piece of code, the
code's execution time is less than a simple system call, if a lock contention
happens, however in current implemenation, the lock holder has to extend its
locking time and enter kernel to unlock it, the change avoids this disadvantage,
it first sets mutex to free state and then enters kernel and wake one waiter
up. This improves performance dramatically in some sysbench mutex tests.
Tested by: kris
Sounds great: jeff
problem where Adaptec's arcconf monitoring tool hangs after producing
its expected output.
Submitted by: Adaptec, via driver ver 15317
MFC after: 1 week
from the softc.
- Rework the watchdog timer to match other NIC drivers:
- Start a timer in fe_init() that runs once a second and checks a counter
in the softc that is identical to the deprecated 'if_timer'.
- Just adjust the softc tx timeout value when sending packets instead of
scheduling the timer.
- Use IFQ_SET_MAXLEN().
Tested by: WATANABE Kazuhiro
FIFO, as required by SUSv3. No specific privilege check is performed
in this case, as FIFOs may be created by unprivileged processes
(subject to the normal file system name space restrictions that may be
in place).
Unlike the Apple implementation, we reject requests to create a FIFO
using mknod(2) if there is a non-zero dev argument to the system call,
which is permitted by the Open Group specification ("... undefined
..."). We might want to revise this if we find it causes
compatibility problems for applications in practice.
PR: kern/74242, kern/68459
Obtained from: Apple, Inc.
MFC after: 3 weeks
performed. Otherwise if ruleset is used by given mountpoint and is empty
it's freed by devfs_ruleset_reap and pointer becomes bogus.
Submitted by: Mateusz Guzik <mjguzik@gmail.com>
PR: kern/124853
some time now so collapse calls accordingly.
o Given that gem_load_txmbuf() is allowed to fail resulting in a packet
drop also for quite some time now implement the functionality of
gem_txcksum() by means of m_pullup(9), which de-obfuscates the code
and allows to always retrieve the correct length of the IP header.
o Add missing BUS_DMASYNC_PREREAD when syncing the control DMA maps in
gem_rint() and gem_start_locked().
o Correct some bus_barrier(9) calls to do a read/write barrier as we
do a read after a write. Add some missing ones in gem_mii_readreg()
and gem_mii_writereg().
o According to the Apple GMAC driver, the GEM ASIC specification and
the OpenSolaris eri(7D) the TX FIFO threshold has to be set to 0x4ff
for the Gigabit variants and 0x100 for the ERI in order do avoid TX
underruns.
o In gem_init_locked():
- be conservative and enable the RX and TX MACs,
- don't clear GEM_LINK otherwise we don't ever mark the link as up
again if gem_init_locked() is called from gem_watchdog(),
- remove superfluous setting of sc_ifflags.
o Don't bother to check whether the interface is running or whether its
queue is empty before calling gem_start_locked() in gem_tint(), the
former will check these anyway.
o Call gem_start_locked() in gem_watchdog() in order to try to get
some more packets going.
o In gem_mii_writereg() after reseting the PCS restore its configuration.
GMAC testing: grehan, marcel
MFC after: 2 weeks
on the amd64 architecture. The amd64 architecture requires kernel code and
global variables to reside in the highest 2GB of the 64-bit virtual address
space. Thus, the memory allocated during bootstrap, before the call to
kmem_init(), starts at KERNBASE, which is not necessarily the same as
VM_MIN_KERNEL_ADDRESS on amd64.
PowerPC/AIM. Consequently, it should not be used to determine the maximum
number of kernel map entries. Intead, use VM_MIN_KERNEL_ADDRESS, which marks
the start of the kernel map on all architectures.
Tested by: marcel@ (PowerPC/AIM)
KERNBASE and VM_MIN_KERNEL_ADDRESS are no longer the same, the physical
memory allocated during bootstrap will be offset from the low-end of the
kernel's page table.
address space on the amd64 architecture. The amd64 architecture
requires kernel code and global variables to reside in the highest 2GB
of the 64-bit virtual address space. Thus, KERNBASE cannot change.
However, KERNBASE is sometimes used as the start of the kernel virtual
address space. Henceforth, VM_MIN_KERNEL_ADDRESS should be used
instead. Since KERNBASE and VM_MIN_KERNEL_ADDRESS are still the same
address, there should be no visible effect from this change (yet).
That said, kris@ has tested crash dumps under the full patch that
increases the kernel virtual address space on amd64 to 6GB.
Tested by: kris@
address space on the amd64 architecture. The amd64 architecture
requires kernel code and global variables to reside in the highest 2GB
of the 64-bit virtual address space. Thus, KERNBASE cannot change.
However, KERNBASE is sometimes used as the start of the kernel virtual
address space. Henceforth, VM_MIN_KERNEL_ADDRESS should be used
instead. Since KERNBASE and VM_MIN_KERNEL_ADDRESS are still the same
address, there should be no visible effect from this change (yet).
This is needed for correct behavior when packets are lost or reordered.
PR: kern/123950
Reviewed by: andre@, silby@
Reported by: Yahoo!, Wang Jin
MFC after: 1 week
needed to promote cdev to cdev_priv, the si_priv pointer was followed.
Use member2struct() to calculate address of the wrapping cdev_priv.
Rename si_priv to __si_reserved.
Tested by: pho
Reviewed by: ed
MFC after: 2 weeks
libi386's time(), caused by a qemu bug. The bug might
be present in other BIOSes, too.
qemu either does not simulate the AT RTC correctly or
has a broken BIOS 1A/02 implementation, and will return
an incorrect value if the RTC is read while it is being
updated.
The effect is worsened by the fact that qemu's INT 15/86
function ("wait" a.k.a. usleep) is non-implmeneted or
broken and returns immediately, causing beastie.4th to
spin in a tight loop calling the "read RTC" function
millions of times, triggering the problem quickly.
Therefore, we keep reading the BIOS value until we get
the same result twice. This change fixes beastie.4th's
countdown under qemu.
Approved by: des (mentor)
- only one functino to destroy an SCTP stack sctp_finish()
- Make it so this function also arranges for any threads
created by the image to do a kthread_exit()
of whether NETATALKDEBUG is enabled, so make building it conditional on
NETATALK instead. This problem appears to have been present from the time
that the netatalk implementation was imported.
PR: 124456
Submitted by: Nathan Whitehorn <whitehorn at wisc dot edu>
MFC after: 3 days
sgtty was the original interface to configure terminal attributes on my
UNIX-like operating systems. It has been deprecated by the POSIX termios
interface, which is implemented in almost any modern system.
An advantage of turning this into a binary compatibility interface, is
that we can now eventually remove the COMPAT_43TTY switch from kernel
configurations. This removes many ioctl()'s from the TTY layer.
While there, increase the __FreeBSD_version, which may be useful for the
people working on the Ports tree.
Reviewed by: kib
Approved by: philip (mentor)
- Vimage prep - these are major restructures to move
all global variables to be accessed via a macro or two.
The variables all go into a single structure.
- Asconf address addition tweaks (add_or_del Interfaces)
- Fix rwnd calcualtion to be more conservative.
- Support SACK_IMMEDIATE flag to skip delayed sack
by demand of peer.
- Comment updates in the sack mapping calculations
- Invarients panic added.
- Pre-support for UDP tunneling (we can do this on
MAC but will need added support from UDP to
get a "pipe" of UDP packets in.
- clear trace buffer sysctl added when local tracing on.
Note the majority of this huge patch is all the vimage prep stuff :-)
same as the global variable defined in ip_input.c. Instead, adopt the name
'q' as found in about 1/2 of uses in ip_input.c, preventing a collision on
the name. This is non-harmful, but means that search and replace on the
global works less well (as in the virtualization work), as well as indexing
tools.
MFC after: 1 week
Reported by: julian
before PG_M. This sometimes prevents unnecessary removal of write access
from a PTE. Overall, the net result is fewer demotions and promotion
failures.
- Add a mutex to the softc to protect the softc and device hardware.
- Use a private watchdog timer.
- Setup interrupt handler after ether_ifattach().
- Use bus_foo() rather than bus_space_foo() and remove bus space tag and
handle from softc.
Tested by: imp
Now that we got rid of the minor-to-unit conversion and the constraints
on device minor numbers, we can convert the functions that operate on
minor and unit numbers to simple macro's. The unit2minor() and
minor2unit() macro's are now no-ops.
The ZFS code als defined a macro named `minor'. Change the ZFS code to
use umajor() and uminor() here, as it is the correct approach to do
this. Also add $FreeBSD$ to keep SVN happy.
Approved by: philip (mentor), pjd
page table page. The direction of the traversal can matter if
pmap_promote_pde() has to remove write access (PG_RW) from a PTE that hasn't
been modified (PG_M). In general, if there are two or more such PTEs to
choose among, it is better to write protect the one nearer the high end of
the page table page rather than the low end. This is because most programs
access memory in an ascending direction. The net result of this change is a
sometimes significant reduction in the number of failed promotion attempts
and the number of pages that are write protected by pmap_promote_pde().
Except for the case where we use the cloner library (clone_create() and
friends), there is no reason to enforce a unique device minor number
policy. There are various drivers in the source tree that allocate unr
pools and such to provide minor numbers, without using them themselves.
Because we still need to support unique device minor numbers for the
cloner library, introduce a new flag called D_NEEDMINOR. All cdevsw's
that are used in combination with the cloner library should be marked
with this flag to make the cloning work.
This means drivers can now freely use si_drv0 to store their own flags
and state, making it effectively the same as si_drv1 and si_drv2. We
still keep the minor() and dev2unit() routines around to make drivers
happy.
The NTFS code also used the minor number in its hash table. We should
not do this anymore. If the si_drv0 field would be changed, it would no
longer end up in the same list.
Approved by: philip (mentor)
mtx interface for NDIS_LOCK/UNLOCK. This should result in less
CPU utilization on behalf of the ndis driver. Additionally, this
commit also fixes a potential LOR in the ndis_tick code, by
not locking inside the ndis_tick function, but instead delegating
that work to the helpers called through IoQueueWorkItem. The
way that this is currently set up for NDIS prevents us from
simply implementing a callout_init_mtx mechanism.
However, the helper functions that handle the various timeout
cases implement fine-grained locking using the spinlocks provided
by the NDIS-compat layer, and using the mtx that is added with
this commit. This leaves the following ndis_softc members operated
on in ndis_tick in an unlocked context:
* ndis_hang_timer - Only modified outside of ndis_tick once, before
the first callout_reset to schedule ndis_tick
* ifp->if_oerrors - Only incremented in two places, which should be
an atomic op
* ndis_tx_timer - Assigned to 5 (when guaranteed to be 0) or 0
(in txeof), to indicate to ndis_tick what to
do. This is the only member of which I was
suspicious for needing the NDIS_LOCK here. My
testing (and another's) have been fine so far.
* ndis_stat_callout - Only uses a simple set of callout routines,
callout_reset only called by ndis_tick after
the initial reset, and then callout_drain is
used exactly once in shutdown code.
The benefit is that ndis_tick doesn't acquire NDIS_LOCK unless one of
the timeout conditions is flagged, and it still obeys the locking
order semantics that are dictated by the NDIS layer at the moment. I
have been investigating a more thorough s/spinlock/mtx/ of the NDIS
layer, but the simplest naive approach (replace KeAcquireSpinLock
with an mtx implementation) has anti-succeeded for me so far. This
is a good first step though.
Tested by: onemda@gmail.com
Reviewed by: current@, jhb, thompsa
Proposed by: jhb
The while loop that is assumed to initialize the uio_off later, may
be not entered at all, causing uninitialized value to be returned in
uio->uio_offset.
PR: 122925
Submitted by: Jaakko Heinonen <jh saunalahti fi>
MFC after: 1 weeks
We still use the interrupt filter due to performance problems that show up if
we don't. The main problem seen is that, due to the interrupt being edge
triggered, we occasionally miss interrupts which leads us to not notice that
we can transmit more packets. Using the new approach, which just schedules
a task on a taskqueue, we are guaranteed to have the task run even if the
interrupt arrived while we were already executing. If we were to use an
ithread the system would mask the interrupt while the handler is run and we'd
miss interrupts.
allocated semaphores, so it's wrong to increase it conditionally,
in this case for every over-the-limit semaphore nsegs is decreased
without being previously increased.
PR: kern/123685
Approved by: cognet (mentor)
- Add a mutex to the softc to protect the softc and device hardware.
- Use a private timer to implement a watchdog for tx timeouts and drive
the timer for auto negotiation.
- Use bus_foo() rather than bus_space_foo() and remove the bus space
tag & handle from the softc.
- Call bus_setup_intr() after ether_ifattach().
Tested by: Florian Smeets flo of kasimir.com
Remove the code which disables port status change interrupts for 1s
when one occured -- this makes that events get lost or delayed until
the next change.
Obtained from: NetBSD
- Fixed a problem on i386 architecture when using split header/jumbo frame
firmware caused by hardware alignment requirements.
- Added #define BCE_USE_SPLIT_HEADER to allow the feature to be enabled/
disabled. Enabled by default.
PR: kern/123696
MFC after: 2 weeks
and nfs requests processing. Lockmgr lock provides the shared locking for
nfs requests, while exclusive mode is used for modifications. The writer
starvation is handled by lockmgr too.
Reported by: kris, pho, many
Based on the submission by: mohan
Tested by: pho
MFC after: 2 weeks
In the FreeBSD base system, there are only two utilities that use struct
tty, namely pstat and sicontrol. The sicontrol utility calls the
TCSI_TTY ioctl(), which copies struct tty back to userspace.
sicontrol should not have this functionality. The same data is already
provided by pstat. If we really want to be able to export these numbers
through a file descriptor to userspace, we can export struct xtty, which
should provide a better abstraction. The ttystat option was only used as
a debugging aid.
This makes sicontrol compile in the mpsafetty branch.
Reviewed by: peter
Approved by: philip (mentor)
newvers.sh is run pwd is actually the obj directory, so "../../.svn"
doesn't exist and the test always fails. The second is that buildkernel
is executed with a restrictive PATH, so unless you have svnversion in
/bin or /usr/bin it can't run.
Fix this by looking for svnversion in /bin, /usr/bin, and /usr/local/bin
in that order. If found, store the location and derive the value of the
source directory. Then run svnversion in the appropriate directory.
There is one possible refinement which would be to add a test for
LOCALBASE!=/usr/local if we don't find svnversion the first time, but
IMO that's not necessary at this time.
is in little endian form. Likewise setting DC_AL_PAR0/DC_AL_PAR1
register expect the address to be in little endian form. For big
endian architectures the address should be swapped to get correct
one.
Change setting/getting ethernet hardware address to big endian
architecture frendly.
Reported by: Robert Murillo ( billypilgrim782001 at yahoo dot com )
Tested by: Robert Murillo ( billypilgrim782001 at yahoo dot com )
some longstanding issues:
o pass the vap since it's now the "coin of the realm" and required
to do things like set initial tx parameters in private node
state for use prior to association
o pass the mac address as cards that maintain outboard station
tables require this to create an entry (e.g. in ibss mode)
o remove the node table reference, we only have one node table
and it's unlikely this will change so this is not needed to
find the com structure
entry in the SMAP is a 20 byte structure and they are queried from the
BIOS via sucessive BIOS calls. Due to an apparent bug in the R900's
BIOS, for some SMAP requests the BIOS overflows the 20 byte buffer
trashing a few bytes of memory immediately after the SMAP structure. As
a workaround, add 8 bytes of padding after the SMAP structure used in
the loader for SMAP queries.
PR: i386/122668
Submitted by: Mike Hibler mike flux.utah.edu, silby
MFC after: 3 days
- Store the softc of the device in the 'si_drv1' of the cdev.
- Lookup the softc via 'si_drv1' in cdev methods rather than using the
minor number as a unit for devclass_get_softc().
- Lookup the device_t via the softc field in cdev methods rather than
using the minor number as a unit for devclass_get_device().
- Add a mutex to the softc to protect 'sc_opened'.
- Remove D_NEEDGIANT as all the smbus drivers are now MPSAFE and this driver
is now MPSAFE.
- Remove some checks for NULL softc pointers that can't happen and don't
bzero the softc during attach.
work. (Moreover, I don't believe that they have ever worked as intended.)
The explanation is fairly simple. Both MADV_DONTNEED and MADV_FREE perform
vm_page_dontneed() on each page within the range given to madvise(). This
function moves the page to the inactive queue. Specifically, if the page is
clean, it is moved to the head of the inactive queue where it is first in
line for processing by the page daemon. On the other hand, if it is dirty,
it is placed at the tail. Let's further examine the case in which the page
is clean. Recall that the page is at the head of the line for processing by
the page daemon. The expectation of vm_page_dontneed()'s author was that
the page would be transferred from the inactive queue to the cache queue by
the page daemon. (Once the page is in the cache queue, it is, in effect,
free, that is, it can be reallocated to a new vm object by vm_page_alloc()
if it isn't reactivated quickly enough by a user of the old vm object.) The
trouble is that nowhere in the execution of either MADV_DONTNEED or
MADV_FREE is either the machine-independent reference flag (PG_REFERENCED)
or the reference bit in any page table entry (PTE) mapping the page cleared.
Consequently, the immediate reaction of the page daemon is to reactivate the
page because it is referenced. In effect, the madvise() was for naught.
The case in which the page was dirty is not too different. Instead of being
laundered, the page is reactivated.
Note: The essential difference between MADV_DONTNEED and MADV_FREE is
that MADV_FREE clears a page's dirty field. So, MADV_FREE is always
executing the clean case above.
This revision changes vm_page_dontneed() to clear both the machine-
independent reference flag (PG_REFERENCED) and the reference bit in all PTEs
mapping the page.
MFC after: 6 weeks
it. Bad imp. Removing us dips us under 10,000 in size too.
o Replace an unconditional 30ms DELAY (yes, busy wait) with a check of the
SIBUSY bit in the SelfST register before accessing the eeprom. This changes
the time to read the EEPROM from 2 * 20 * 30ms (1.2s) to < 20*25us (.0005s)
and make the attach of the card tolerable when ethernet media is present.
Include data from the datasheet about why this works. While this is a 2500x
speed increase, it doesn't really matter at all once the card is probed...
o set dev earlier in softc.
o remove unused fields from softc and args from cs_alloc_irq
o remove some commented code that will never be implemented.
o Don't try to send a packet and see if it worked. We don't
need this anymore, and it doesn't add any value.
o tweaks for BNC and AUI.
o limit possible time hung in the kernel to 4s rather than 40s.
boards. This is enough to net-boot to multiuser.
Also supported is the SMSC LAN91C111 parts used on the netCF, netDUO and netMMC
add-on boards.
I'll be putting some instructions on how to boot this on the Gumstix boards
online soon.
This is still fairly rough and will be refined over time but I felt it was
better to get this out there where other people can help out.
sn(4) driver and also looking at newer drivers. The reason for the rewrite is
to support MII and to try and resolve some performance issues found when trying
to use the sn(4) driver on the Gumstix network boards.
For reference, the SMSC LAN91C111 is a non-PCI ethernet part whose lineage
dates back to Ye Olde Days of ISA. It seems to get some use in the embedded
space these days on parts lacking on-board MACs or on-board PCI controllers,
such as the XScale PXA line of ARM CPUs.
This also includes a driver for the SMSC LAN83C183 10/100 PHY.
Man page to follow.
they can re-added. Remove CS_NAME. Don't whine when there's an
ignored checksum error: User has said STFU, so we should S the FU.
(remove mandated properties).
The CTRL() macro seems to perform character to control-character
conversion (i.e. 'A' to 0x01) to lowercase characters. This is actually
not valid. If we use lowercase characters, conversions such as
CTRL('\\') and CTRL('?') will result to invalid conversions.
Because we must still support old source code that uses CTRL() (bad!),
we make CTRL() accept both forms. When the character is a lowercase
character, we perform the old style conversion.
Approved by: philip (mentor)
- Add a mutex to the softc to protect the softc and the device hardware.
- Add a private timer to manage transmit watchdogs rather than using
if_timer/if_watchdog.
- Setup the interrupt handler after ether_ifattach().
Tested by: imp
for this driver is called 'ie'. Otherwise, ifconfig(8) doesn't recognize
any of the modules as being the ie(4) driver and will always try to kldload
the driver even when it is already present in the kernel.
Reported by: Thierry Herbelot
the dm_lock is held while the newly allocated vnode is locked. Since no
other threads may try to lock the new vnode yet, the LOR there cannot
result in the deadlock.
Shut down the witness warning to note this fact.
Tested by: pho
Prodded by: attilio
10BaseT' since it required 10BaseT to have carrier to switch to it.
This chip makes it hard to do proper auto, so we don't do it. We
can't test carrier on things easily.
Don't insist on carrier when we set the media. Don't report failures.
Remove a 1s! delay that appears to not be needed.
With these patches, and John Baldwin's patches, I'm able to pass
packets on my IBM EtherJet card again.
MAC events.
- Use bus_*() rather than bus_space_*() and remove the bus space tag and
handle from the softc.
- Retire unused macros for examining CIS tuples.
o When forced to be 10baseT, don't require that the 10baseT interface
have link to succeed. Still require it for IFM_AUTO, however, since it
appears that there's no way to tell if a specific type of interface
worked. I'm doing a web search for a datasheet now to see if there's
anything obvious.
o Minor incidental formatting nits, including collapsing code of the form
if (foo) {
bar();
} else {
if (baz)
bing();
}
into:
if (foo) {
bar();
} else if (baz) {
bing();
}
to save an indentation level.
o Remove stray reference to 3.x config file syntax.
# I believe John's patches still apply after this...
timer by keeping a once-a-second timer running that decrements a counter
similar to if_timer and reset the chip if it gets down to zero via the
decrement.
- Use IFQ_SET_MAXLEN().
The Giant lock is acquired in two places in tty_tty.c. In both places,
it is unneeded.
There is no reason to specify D_NEEDGIANT on this device node. The
device node has only been designed to return ENXIO when opened. It
doesn't make any sense to lock/unlock Giant, just to return this error.
D_TTY is also unneeded. The unimplemented functions don't need to be
patched by devfs.
We don't need to lock Giant when we want to lookup the proper TTY vnode.
s_ttyvp is already protected by proctree_lock (see devfs_vnops.c).
Approved by: philip (mentor)
systems where the CardBus bridge was connected to a APIC. The case
where the probe routine is told to not setup the IRQ was mishandled
but the error was masked in the case where the IRQ was a valid one
for the card.
MFC after: 1 week
bring it more up to date. The watchdog timer, and its
associated code, is all collapsed into the ndis_tick function
that was implemented for the NDIS-subsystem watchdog. This
implementation is similar to what numerous other drivers use
to implement the watchdog.
Reviewed by: thompsa, jhb
MFC after: 2 weeks
- Add a mutex to the softc to protect the softc and device hardware.
- Don't leak bus resources if if_alloc() fails during attach.
- Setup the interrupt handler after calling ether_ifattach().
- Use a private timer to manage the transmit watchdog.
Tested by: WATANABE Kazuhiro CQG00620 of nifty.ne.jp
- Add a mutex to protect the softc and device hardware.
- Use a callout rather than a callout_handle for the media timer.
- Use a dedicated timer for managing the tx watchdog rather than if_timer.
- Fix some resource leaks if xe_attach() fails.
- Shutdown the device before detaching the driver.
- Setup the interrupt handler after ether_ifattach().
Tested by: Ian FREISLICH ianf of clue.co.za
- Add a mutex to the softc and use it to protect the softc and device.
- Setup the interrupt handler in the common code instead of in each front
end and do it after ether_ifattach().
- Use ie_stop() and ieinit_locked() in iereset() rather than frobbing IFF_UP
and invoking ieioctl().
- Use DELAY() to implement a spin loop on a register with a timeout rather
than scheduling a timeout and then doing a tight spin on the register.
In the non-MPSAFE case this would never have worked because the spinning
code held Giant and the timeout routine would have been blocked on Giant
forever. The same approach would not worke in the MPSAFE case either for
the same reason, hence use a loop around DELAY().
- Clear IFF_DRV_(RUNNING|OACTIVE) in ie_stop() rather than in callers.
- Call ieinit_locked() directly rather than ieioctl(!) from ie_mc_reset().
- Don't leak the rx frame buffer on detach.
Tested by: Thierry Herbelot thierry of herbelot.com
a client reboot, do this check before performing the lock otherwise we
will trash the new lock along with any other old locks the client held
before rebooting.
Make sure nlm_check_idle always returns with nlm_global_lock held.
MFC after: 1 week
template, use an M_TEMP malloc(9) allocation rather than an mbuf
with mtod(9) and dtom(9). This eliminates the last use of
dtom(9) in TCP.
MFC after: 3 weeks
In the mpsafetty branch, Linux sshd seems to work properly inside a
jail. Some small modifications had to be made to the Linux compatibility
layer.
The Linux PTY routines always expect the device major number to be 136
or higher. Our code always set the major/minor number pair to 136:0.
This makes routines like ttyname() and ptsname() fail, because we'll end
up having ambiguous device numbers.
The conversion was not performed on all *stat() routines, which meant in
some cases the numbers didn't get transformed. By pushing the conversion
into linux_driver_get_major_minor(), the transformation will take place
on all calls.
Approved by: philip (mentor), rdivacky
to reduce performance degradation under heavy outgoing scan/flood.
Scalability is now much more important then several kilobytes of RAM.
Remove unneded TCP-specific expiration handeling. Before this connected
TCP sessions could never expire. Now connected TCP sessions will expire
after 24hours of inactivity.
Simplify HouseKeeping() to avoid several mul/div-s per packet. Taking into
account increased LINK_TABLE_OUT_SIZE, precision is still much more then
required.
- to increase performance do not reallocate mbuf when possible,
- to support up to 16K packets (was 2K max) use mbuf cluster of proper size.
This change depends on recent ng_nat and ip_fw_nat changes.
As discussed with Robert Watson and John Baldwin, it would be better if
PTY's are created with proper permissions, turning grantpt() into a
no-op.
Bypassing security frameworks like MAC by passing NOCRED to
VOP_SETATTR() will only make things more complex.
Approved by: philip (mentor)
pretend to be IntelliMouse (which have a few more features than generic mice)
causing the IntelliMouse probe to work and the Synaptics code never to be
called.
This should not break "real" IntelliMouse because the Synaptics detection code
is fairly specific.
PR: kern/120833
Submitted by: Eygene Ryabinkin <rea-fbsd -at- codelabs.ru>
MFC after: 1 week
This fixes packet fragmentation handeling.
Pass really available buffer size to libalias instead of MCLBYTES constant.
MCLBYTES constant were used with believe that m_megapullup() always moves
date into a fresh cluster that sometimes may become not so.
promotion within the kernel's address space. Specifically,
pmap_promote_pde() is only called when the page table page (PTP) that
is referenced by the given PDE has a full "use count", i.e., its
wire_count is 512. Although this guarantees for a user address space
that all 512 PTEs in the PTP hold valid mappings, the same is not true
of the kernel's address space. A kernel PTP always has a use count of
512 regardless of the state of the PTEs. Therefore,
pmap_promote_pde() should not assume (or assert) that the first PTE in
the PTP is valid.
(Don't ask for a vendor import of this yet, we're in the early days of svn)
Instead of using cyclic timers to call the state clean and deadman callbacks,
use a callout on FreeBSD to avoid the deadlock on FreeBSD due to trying to
send interprocessor interrupts with interrupts disabled.
Reported by: ps, jhb, peter, thompsa
In the mpsafetty branch, PTY's are allocated through the posix_openpt()
system call. The controller side of a PTY now uses its own file
descriptor type (just like sockets, vnodes, pipes, etc).
To remain compatible with existing FreeBSD and Linux C libraries, we can
still create PTY's by opening /dev/ptmx or /dev/ptyXX. These nodes
implement d_fdopen(). Devfs has been slightly changed here, to allow
finit() to be called from d_fdopen().
The routine grantpt() has also been moved into the kernel. This routine
is a little odd, because it needs to bypass standard UNIX permissions.
It needs to change the owner/group/mode of the slave device node, which
may often not be possible. The old implementation solved this by
spawning a setuid utility.
When VOP_SETATTR() is called with NOCRED, devfs_setattr() dereferences
ap->a_cred, causing a kernel panic. Change the de_{uid,gid,mode} code to
allow changes when a->a_cred is set to NOCRED.
Approved by: philip (mentor)
whatever frequency it started at instead of always picking the highest
frequency. The first version of this driver attempted to do this, but it
set the speed to the first frequency in the list rather than the value it
had saved.
MFC after: 1 week
Discussed with: rpaulo, phk
clients that have rebooted (or otherwise changed port numbers). If the
client is broken or has no active locks, it won't notify us. Fall back
on the two minute timeout logic used by the userland rpc.lockd code.
MFC after: 1 week
variations from normal 16x50 behaviour however is the the use of a normally
unused bit of IER to control RX timeout interrupts independently of the
generally used RXRDY bit. If this bit is not enabled, we only ever get
interrupts when the FIFO is full, never before. This is not very useful when
the UART is being used as a console.
In order to support this without causing potential problems on more "normal"
16x50 variants, this change introduces two hints for the uart device, ier_mask
and ier_rxbits. These can be used to override which bits get set and cleared
when we're enabling and disabling RX interrupts.
Reviewed by: marcel
some cases, add explicit inpcb locking rather than relying on the global
lock, as we dereference inp_socket, but also allowing us to drop the
global lock more quickly.
MFC after: 1 week
Even though we got rid of device major numbers some time ago, device
drivers still need to provide unique device minor numbers to make_dev().
These numbers are only used inside the kernel. They are not related to
device major and minor numbers which are visible in devfs. These are
actually based on the inode number of the device.
It would eventually be nice to remove minor numbers entirely, but we
don't want to be too agressive here.
Because the 8-15 bits of the device number field (si_drv0) are still
reserved for the major number, there is no 1:1 mapping of the device
minor and unit numbers. Because this is now unused, remove the
restrictions on these numbers.
The MAXMAJOR definition was actually used for two purposes. It was used
to convert both the userspace and kernelspace device numbers to their
major/minor pair, which is why it is now named UMINORMASK.
minor2unit() and unit2minor() have now become useless. Both minor() and
dev2unit() now serve the same purpose. We should eventually remove some
of them, at least turning them into macro's. If devfs would become
completely minor number unaware, we could consider using si_drv0 directly,
just like si_drv1 and si_drv2.
Approved by: philip (mentor)
monitoring UDP connections using sysctls. In some cases, add
previously missing locking of inpcbs, as inp_socket is followed,
which also allows us to drop global locks more quickly.
MFC after: 1 week
clocked at 10x normal speed. That is, when you set it for 9600
baud, it actually does 96000 baud. In order to make it plug and
play with other serial ports, it has to have its clock rate
reduced by a factor of 10.
Discussed with: Marcel Moolenaar
MFC after: 2 weeks
o do not put the chip into full sleep in ath_stop as it gains
nothing and causes many parts to hang in ath_detach because we
may touch the chip during vap teardown; this may also fix issues
with unloading the module
o add a note in ath_detach to explain ath_hal_detach puts the
chip in low power mode; this is useful to know as it means
unloading the module will place a pci device in the lowest
possible power state
o leave an #ifdef notyet marker for powering down the chip when
a device is marked down; we can't do that until we handle all
the ways the driver may be entered and touch the chip
o fix resume by reloading the h/w key cache as it's been clobbered
(for pci) by the socket being powered off; for station mode we
directly stop+init the chip and then simulate a beacon miss to
get the upper layers sync'd up; for other configs we must brute
force stop+start the vaps so they go through the state machine
address specified in the ioctl and for drivers that need the address
to locate a key (e.g. for delete).
Note this changes net80211-private api's but not the driver callback;
may want to change that in the future.
Reviewed by: sephe, thompsa
on amd64. Note the only difference is the iovec32 part so I use the
native structure for everything else.
Also I plan to MFC all the changes in -current to 7-stable and 6-stable
shortly since I've been running them. This does not include the cam
changes.
MFC after: 3 days
o construct a name for the com lock as done for other locks
o pass the device name to IEEE80211_LOCK_INIT so the mtx name
is constructed as foo_com_lock
o introduce *_LOCK_OBJ macro's to hide the lock contents and
minimize redundant code