When reading the code I had to stop, say "ok, what does *these*
modifications of strl*() do? Pull out grep. Oh, not in add/, maybe above
in ../lib/? Yep. So what do they do? Comments above them are misleading,
guess I'll have to read the code. Oh, they just test strl* against the
size and return the result of the test. Now I can continue to read the
code I was.
The uses of s_strl*() then test that result and errx()'s.
Lets think about the "optimized" code I am removing:
In general the compiler pushes the three args to strl* onto the stack and calls
s_strl*. s_strl* has to indirectly access 3 args from the stack. Then push
them on the stack a 2nd time for the real strl* call. s_strl* then pops the
return from strl* off the stack; or moves it from the register it was returned
in, to the register where tests can happen. s_strl* then pops the three
arguments to strl*. Perform the test, push the result of the test, or move it
from the result register to the return value register. The caller to s_strl*
now has to either pop the return value of s_strl* or move it from the return
value register to the test register. The caller then pops the three args to
s_strl* off the stack (the same args that s_strl* itself had to pop off after
the real call to strl*). The s_strl* caller then performs a simular test to
what has already been done, and conditionally jumps. By doing things this way, we've given the compiler optimizer less to work with.
Also, please don't forget the that call to s_strl* has possibly jumped to code
not in the cache due to being far away from the calling code, thus causing a
pipeline stall.
So where is the "optimization" from s_strl*?
It isn't code clarity.
It isn't code execution speed. It isn't code size either.
never completed" message. The RX reset takes longer complete than it
used to, a lot longer in fact than xl_wait() is prepared to wait.
When we do the RX reset in xl_reset(), this cases xl_wait() to time out
and whine. We wait a little extra time now after the RX reset, which
should silence the warning.
Thanks to obrien for finally getting me a box with a NIC that
causes this problem for me to tinker with.
entry (d_ino == 0) is found in a position that is not the start of
a DIRBLKSIZ block.
While such entries cannot occur normally (ufs always extends the
previous entry to cover the free space instead), they do not cause
problems and fsck does not fix them, so panicking is bad.
us anyway because it doesn't work right on the x86 and alpha. On
K&R code, small ints would be promoted to int. ANSI-C doesn't require
this and the small ints can be passed taking 8 or 16 bits of stack
space. However, the x86 abi that we use *does* promote to 32 bit,
and the alpha ABI passes them in 64 bit registers so we dont have
that aspect of the problem here. Losing float precision by having it
cast down to int because the funtion prototype specifies int is the
least of our problems. -Wmissing-prototypes helps here anyway.
hw.pcic.irq Globally set the IRQ for all pcic devices' management
interrupt (aka card status change or CSC interrupt)
This is what used to be known as
machdep.pccard.pcic_irq (which has been retained for
now for compatibility).
hw.pcic.ignore_fuction_1 Ignores function 1 for all PCIC bridges by not
attaching to them. Lucent released a huge batch
of cards that were imporperly manufactuered (lacking
the 0 ohm resister to disable slot 1). This is
a big hammer to keep those cards from causing problems
(I've had 4 people contact me saying my patches
worked great once they added a kludge to always ignore
function 1, or until they soldered these resistors
in place!).
No clue where to document these. They act as both boot loader environment
variables, as well as read-only sysctls after boot.
At the same time, sort sys/systm.h in its proper order after sys/sysctl.h.
VM caching of disks through mmap() and stopping syncing of open files
that had their last reference in the fs removed (ie: their unsync'ed
pages get discarded on close already, so I made it stop syncing too).
o kill blank line that I introduced in cardinfo.h
o Delete unused variable wasinactive.
o return 0 from pccard_resume.
o Set the state and lastsate initially to be empty.
o move comment above code for interrupt dispatching.
o Powerstate interface is now available as of 430002, not 500000 (note that
this change will be not 100% correct since the power state stuff didn't
enter current until well after 500000, but it is good enough for the two
branche we have going now).
power x 0.
pccardc power x 0 used to disable the slot. But a suspend/resume
would reactivate the pccard. It no longer does that. Now the
disabling of the slot is sticy until it is reset with power x 1 or the
card is ejected. This seems closer to correct behavior to me.
o Process all card state changes the same using pccard_do_stat_change().
o Cleanup disabling the card so that we can preserve the state after
the change. Basically, don't set it to empty as often as we do.
o On suspend, the new state is "empty" and the laststate is "suspend"
o Document state machine with a diagram of states and edges. The
edges are labeld to tell the reader what event causes the external
state changes.
o "machdep.pccard.pcic_resume_reset" may be obsolete now. We always
call the bridge driver's resume method on resume now. Otherwise cards
won't automatically show up. If it needs to stay, I'll add it back.
In file included from ../../../dev/vinum/vinumhdr.h:77,
from ../../../dev/vinum/vinum.c:44:
../../../dev/vinum/vinumext.h:165: warning: redundant redeclaration of `setjmp' in same scope
../../../sys/systm.h:96: warning: previous declaration of `setjmp'
../../../dev/vinum/vinummemory.c:44: warning: redundant redeclaration of `longjmp' in same scope
../../../sys/systm.h:97: warning: previous declaration of `longjmp'
longer have a pccard in the slot. This fixes the problem where pccard
would say that a card had been inserted on resume. This also appears
to make the insert/remove events more reliable after a resume as well,
but that may be a different bug I need to hunt down.
initialize in the right order to make derivative settings work right.
eg: at compile time, nmbufs was double nmbclusters. For POLA this should
work the same at runtime.
Tunables are now derived at boot time from maxusers. ie: change maxusers
via a tunable and all the derivative settings change. You can change
the other tunables individually as well. Even hz etc is tunable.
net.link.ether.bridge_refresh variables. While I'm here, try to
make some of the markup on this page more consistent with the
new (markup-reviewed) content.
PR: 22060
Reviewed by: ru (for markup, on an earlier version of this delta)
MFC after: 2 days
into addresses as we have no idea what address family they belong to.
When -n is not specified, resolve IPv6 as well as IPv4 addresses found
in the host field of utmp. Use realhostname_sa() to resolve addresses
(the old code was wrong).
Rename ``x'' to ``x_suffix'' to avoid confusion.
Hard code the host column width to 16 (against the imminent increase
of UT_HOSTSIZE in utmp.h).
making pcbs available to the outside world. otherwise, we will see
inpcb without ipsec security policy attached (-> panic() in ipsec.c).
Obtained from: KAME
MFC after: 3 days
were indices in a dense array. The cpuids are a sparse set and treat
them as such, setting up containers only for CPUs activated during
mb_init().
- Fix netstat(1) and systat(1) to treat the per-CPU stats area as a sparse
map, in accordance with the above.
This allows us to properly boot with certain CPUs disactivated. However, if
we later decide to re-activate said CPUs, we will barf until we decide to
implement CPU spinon/spinoff callback hooks to allow for said CPUs' per-CPU
containers to get configured on their activation.
Reported by: mjacob
Partially (sys/ diffs) Submitted by: mjacob
