kernel: Add m_extadd().

Move static & inline to the beginnings of declarations.

kernel - Adjust mbuf cached free object retention

* Adjust the cache retention in the objcache_create() calls the mbuf
subsystem makes to reduce burst ENOBUF failures. Previously the
main mbufp

kernel - Adjust mbuf cached free object retention

* Adjust the cache retention in the objcache_create() calls the mbuf
subsystem makes to reduce burst ENOBUF failures. Previously the
main mbufphdr cache only retained ~150 or so free mbufs. Now it
retains up to 1/4 the limit.

Reported-by: Peter Avalos <peter@theshell.com>

show more ...

mbuf - Add m_apply and m_getptr

* Import m_apply and m_getptr from FreeBSD.

Obtained-from: FreeBSD

MBUF - Account for failed allocations so netstat -m reports properly.

AMD64 - Refactor uio_resid and size_t assumptions.

* uio_resid changed from int to size_t (size_t == unsigned long equivalent).

* size_t assumptions in most kernel code has been refactored to opera

AMD64 - Refactor uio_resid and size_t assumptions.

* uio_resid changed from int to size_t (size_t == unsigned long equivalent).

* size_t assumptions in most kernel code has been refactored to operate in a
64 bit environment.

* In addition, the 2G limitation for VM related system calls such as mmap()
has been removed in 32 bit environments. Note however that because
read() and write() return ssize_t, these functions are still limited
to a 2G byte count in 32 bit environments.

show more ...

Generally use NULL instead of explicitly casting 0 to some pointer type.

mbuf: Don't discount the objcache limits required by non-pkthdr mbuf

kill db_print_backtrace()

Use print_backtrace() which is always available. Remove
a bunch of #ifdef DDB and #include <ddb/ddb.h> lines.

Since kmalloc's limit testing is not synchronized across CPUs, we raise
the limit by another 1/8 to take the loosememuse into account.

- Return the real cluster limit used by the objcache
- For mbuf objcaches, raise backing kmalloc pools' limit according to the
cluster limits.
Suggested-by: dillon@

Reviewed-by: aggelos@, nth@

Use atomic_fetchadd_int() to simplify the reference counting logic.

Reviewed-by: dillon@

Add m_devpad() to avoid code duplication in various network device drivers

* Fix some cases where NULL was used but 0 was meant (and vice versa).

* Remove some bogus casts of NULL to (void *).

Make m_mclfree() MP safe by fixing a N->0 dereferencing race. The spinlock
wasn't safe because the structure is dynamically allocated and freed.
Instead just use atomic_cmpset_int() to detect the 1-

Make m_mclfree() MP safe by fixing a N->0 dereferencing race. The spinlock
wasn't safe because the structure is dynamically allocated and freed.
Instead just use atomic_cmpset_int() to detect the 1->0 transition.

show more ...

Make mbuf allocator statistics SMP safe.

Make mbtypes[] and mbstat per-cpu. Use atomic_*_*_nonlocked() to adjust
their values. Convert sysctls (kern.ipc.mbstat and kern.ipc.mbtypes) into
procedure

Make mbuf allocator statistics SMP safe.

Make mbtypes[] and mbstat per-cpu. Use atomic_*_*_nonlocked() to adjust
their values. Convert sysctls (kern.ipc.mbstat and kern.ipc.mbtypes) into
procedures that sum the statistics for each CPU.

Reviewed-By: Matt Dillon <dillon@backplane.com>

show more ...

Add the MBUF_DEBUG option. This is a fairly invasive option that should
not be used on production boxes. All active mbufs are tracked with a separate
tracking structure and arranged in a RB tree.

Add the MBUF_DEBUG option. This is a fairly invasive option that should
not be used on production boxes. All active mbufs are tracked with a separate
tracking structure and arranged in a RB tree. A kernel built with this
option can dump all active mbufs with 'sysctl kern.ipc.showmbufs'.

A person debugging the kernel can sprinkle mbuftrackid() calls in the code
to associate an integer id with an mbuf to make the sysctl output more
useful.

The size of the mbuf structure is NOT changed through the use of this option,
but mbuf management overhead will be increased.

show more ...

Make the kern.ipc.nmbclusters and kern.ipc.nmbufs sysctls read-only.
These may only be set as boot-time variables in /boot/loader.conf.

Adjusting them after the fact w/sysctl will not update the obj

Make the kern.ipc.nmbclusters and kern.ipc.nmbufs sysctls read-only.
These may only be set as boot-time variables in /boot/loader.conf.

Adjusting them after the fact w/sysctl will not update the objcache limits
and will cause a ton of confusion because then netstat -m will report
the adjusted limits but the system won't actually be using the adjusted
limits.

objcache mbuf limits can be adjusted dynamically, we just need some code to
do it -> for the future!

Reported-by: Peter Avalos <pavalos@theshell.com>

show more ...

Revamp SYSINIT ordering. Relabel sysinit IDs (SI_* in sys/kernel.h) to
make them less confusing, particularly with regard to the relative order
init routines are called in.

Reorder many sysinits.

Revamp SYSINIT ordering. Relabel sysinit IDs (SI_* in sys/kernel.h) to
make them less confusing, particularly with regard to the relative order
init routines are called in.

Reorder many sysinits. Reorder the SMP and CLOCK code to bring all the
cpus up far earlier in the boot sequence and to make the full threading and
clocking subsystems available for device config.

show more ...

Remove weird license clause which has expired.

Rename printf -> kprintf in sys/ and add some defines where necessary
(files which are used in userland, too).

Pass NULL to objcache_create() to indicate that null_ctor() and/or null_dtor()
should be used. Make null_ctor() and null_dtor() private functions.

Disallow M_ZERO entirely, it doesn't really fit th

Pass NULL to objcache_create() to indicate that null_ctor() and/or null_dtor()
should be used. Make null_ctor() and null_dtor() private functions.

Disallow M_ZERO entirely, it doesn't really fit the objcache model.

With-input-from: "Simon 'corecode' Schubert" <corecode@fs.ei.tum.de>,
"Thomas E. Spanjaard" <tgen@netphreax.net>,
Jeffrey Hsu <hsu@crater.dragonflybsd.org>

show more ...

Rename malloc->kmalloc, free->kfree, and realloc->krealloc. Pass 1

Rename functions to avoid conflicts with libc.

Switch the type and how argument declarations around to match what the
rest of the kernel expects. Note that only the NCP protocol uses m_getm().

Submitted-by: Alexey Slynko <slynko@tronet.ru>