The C code assumes that the carry bit is always kept from the previous
operation. However, the pointer indexing requires another add operation.
Thus, the carry bit from the first operation is tromped over by the
"addl" operation that ends up following it, so the "adcl" that follows
that has no effect because the carry bit is cleared before it.
The result is checksum failure on received packets.
The larger issue is that there isn't any other way of preventing the compiler
inserting arbitrary instructions between different __asm statements (and
that the commit message in revision 1.13 of in_cksum.h is wrong on
this point). From
http://developer.apple.com/documentation/DeveloperTools/gcc-3.3/gcc/Extended-Asm.html
---8<---8<---8<---
You can't expect a sequence of volatile asm instructions to remain
perfectly consecutive. If you want consecutive output, use a single
asm. Also, GCC will perform some optimizations across a volatile
asm instruction; GCC does not "forget everything" when it encounters
a volatile asm instruction the way some other compilers do.
---8<---8<---8<---
Also, this change also makes the ASM code much easier to read.
PR: 69257
Submitted by: Mike Bristow <mike@urgle.com>, Qing Li <qing.li@bluecoat.com>
prevent leakage of Giant. With INVARIANTS, this results in an
assertion failure following execution of the RPC. Without INVARIANTS,
it could result in problems if the NFS server is killed causing nfsd
to return to user space holding Giant.
Feet provided by: brueffer
On a system with huge number of pipes, M_NOWAIT failes almost always,
because of memory fragmentation.
My fix is different than the patch proposed by Pawel Malachowski,
because in FreeBSD 5.x we cannot sleep while holding dummynet mutex
(in 4.x there is no such lock).
My fix is also ugly, but there is no easy way to prepare nice and clean fix.
PR: kern/46557
Submitted by: Eugene Grosbein <eugen@grosbein.pp.ru>
Reviewed by: mlaier
of the MCOUNT_ENTER, MCOUNT_EXIT and MCOUNT_DECL defines. Also make
sure there's a prototype of _MCOUNT_DECL(). This allows us to build
a kernel. There are still unresolved symbols, so linking fails.
_mcount() stub when profiling is enabled. Emit this code sequence
for assembly routines as welli (MCOUNT definition in <machine/asm.h>.
We do not pass the GOT entry however as the 4th argument, because it's
not used. The _mcount() stub calls __mcount(), which does the actual
work. Define _MCOUNT_DECL to define __mcount. We do not have an
implementation of mcount(), so we define MCOUNT as empty, but have a
weak alias to _mcount() in _mcount.S.
Note that the _mcount() stub in the kernel is slightly different from
the stub in userland. This is because we do not have to worry about
nested routines in the kernel.
by default. As such, mark if_dc as IFF_NEEDSGIANT until such
time as appropriate locking review and testing can take place,
and the locking can be enabled by default.
RELENG_5 candidate.
send routine. In IPv6 UDP, the thread will be passed to suser(), which
asserts that if a thread is used for a super user check, it be
curthread. Many of these protocol entry points probably need to
accept credentials instead of threads.
MT5 candidate.
Noticed/tested by: kuriyama
directly. This removes a few more users of the stackgap and also marks
the syscalls using these wrappers MP safe where appropriate.
Tested on: i386 with linux acroread5
Compiled on: i386, alpha LINT
date: 2004/08/22 14:48:55; author: rwatson; state: Exp; lines: +0 -2
Don't need to assert Giant in fw_output(), only in the firewire start
routine.
Approved by: re (scottl)
buffers with kqueue filters is no longer required: the kqueue framework
will guarantee that the mutex is held on entering the filter, either
due to a call from the socket code already holding the mutex, or by
explicitly acquiring it. This removes the last of the conditional
socket locking.
Set the DMA SGL length correctly if the DMA request must be chained because
it is too large to fit in one SGL.
This should fix this driver for some Dell Precision systems.
RELENG_5 candidate.
PR: kern/66479
Submitted by: HITOSHI Osada <qfh02545@nifty.com>
We were obtaining different spin mutexes (which disable interrupts after
aquisition) and spin waiting for delivery. For example, KSE processes
do LDT operations which use smp_rendezvous, while other parts of the
system are doing things like tlb shootdowns with a different mutex.
This patch uses the common smp_rendezvous mutex for all MD home-grown
IPIs that spinwait for delivery. Having the single mutex means that
the spinloop to aquire it will enable interrupts periodically, thus
avoiding the cross-ipi deadlock.
Obtained from: dwhite, alc
Reviewed by: jhb
the geometry code to grab a mutex that prohibits any driver on the
stack below it from sleeping, it's not safe to allow anything in
the top half of isp to sleep (excepting the thread that Fibre Channel
instances use to re-scan loops/fabrics).
hold its own values, pass them up to the parent (acpi0) and merge/uniq them
on the way. After the namespace evaluation, acpi will reserve these
resources and manage them via rman before bus_generic_probe() and
bus_generic_attach(). This is necessary because some systems specify
conflicting resources in separate sysresource objects. It's also cleaner
in that the interface between sysresource and acpi is now merely the parent's
resource list. This code handles the following cases:
1. Unique resource: add it to the parent via bus_set_resource().
2. New wholly contained in old: discard new.
3. New tail overlaps old head: grow old head downward.
AND/OR
4. New head overlaps old tail: grow old tail upward.
Tested by: Pawel Worach <sajd_at_telia.com>
Tested by: Radek Kozlowski <radek_at_raadradd.com>
MFC after: 5 days
with VmWare 4.x. At least with VmWare version 4.5.2, i386 version of
atomic_cmpset_int() is about 30 times slower than non-i386 version. It
makes this delta a good 5.3 MFC candidate, since otherwise it will
mislead users who run FreeBSD under modern VmWare otherwise.
valid; otherwise a caller could trick us into changing any 32-bit word
in kernel memory to LINUX_SOL_SOCKET (0x00000001) if its previous value
is SOL_SOCKET (0x0000ffff).
MFC after: 3 days
The prefix management code currently resides in nd6, leaving only the
unused router renumbering capability in the in6_prefix files. Removing
it will make it easier for us to provide locking for the remainder of
IPv6 by reducing the number of objects requiring synchronized access.
This functionality has also been removed from NetBSD and OpenBSD.
Submitted by: George Neville-Neil <gnn at neville-neil.com>
Discussed with/approved by: suz, keiichi at kame.net, core at kame.net
deal with 24-bit addresses. While the two other attachments, namely
isa and cbus, do it properly, the PCI attachment was passing
BUS_SPACE_MAXADDR instead of BUS_SPACE_MAXADDR_24BIT. This bug
became apparent with the new contigmalloc() code.
This fixes the problem reported with lnc(4) interfaces inside VMWare,
and should theoritically also fix any user of a PCI lnc(4) card. It
is a RELENG_5 MFC candidate.
Tested by: Florian Le Goff <madflo@beertech.org>
position that is 64-bit aligned and makes sure that the valid and
dirty fields are also 64-bit aligned. This means that if PAGE_SIZE
is 32K, the size of the vm_page structure is only increased by 8
bytes instead of 16 bytes. More importantly, the vm_page structure
is either 120 or 128 bytes on ia64. These are "interesting" sizes.
before returning. Device nodes are created via the "taste" mechanism,
so this is necessary in order to make sure that devfs entries are
created before mdconfig(8) returns.
This may be a MFC candidate for 5.3.
Suggested by: phk
allocation. Notably, in this case, the driver tries to allocate several
pieces of memory and then fails if the pieces allocated after the first
do not come after it physically, and within a specific range (8MB I
believe). Of course, this could just as easily fail for any number of
reasons, but it almost always fails now that contiguous allocations start
at the end of possible specified memory locations rather than the beginning.
Allocate all the possibly-needed memory up front, even though it's a waste,
to get around this. The least bogus solution would be to take the physical
address from the first allocation and create a new tag that specified that
further allocations must follow it within that 8MB window, then use that
when allocating new channels, but that's left for anyone else that really
feels like doing it.
Tested by: Erwin Lansing <erwin@lansing.dk>
