routine wakes up proc0 so that proc0 can swap the thread back in.
Historically, this has been done by waking up proc0 directly from
setrunnable() itself via a wakeup(). When waking up a sleeping thread
that was swapped out (the usual case when waking proc0 since only sleeping
threads are eligible to be swapped out), this resulted in a bit of
recursion (e.g. wakeup() -> setrunnable() -> wakeup()).
With sleep queues having separate locks in 6.x and later, this caused a
spin lock LOR (sleepq lock -> sched_lock/thread lock -> sleepq lock).
An attempt was made to fix this in 7.0 by making the proc0 wakeup use
the ithread mechanism for doing the wakeup. However, this required
grabbing proc0's thread lock to perform the wakeup. If proc0 was asleep
elsewhere in the kernel (e.g. waiting for disk I/O), then this degenerated
into the same LOR since the thread lock would be some other sleepq lock.
Fix this by deferring the wakeup of the swapper until after the sleepq
lock held by the upper layer has been locked. The setrunnable() routine
now returns a boolean value to indicate whether or not proc0 needs to be
woken up. The end result is that consumers of the sleepq API such as
*sleep/wakeup, condition variables, sx locks, and lockmgr, have to wakeup
proc0 if they get a non-zero return value from sleepq_abort(),
sleepq_broadcast(), or sleepq_signal().
Discussed with: jeff
Glanced at by: sam
Tested by: Jurgen Weber jurgen - ish com au
MFC after: 2 weeks
for a PCI device during the boot-time probe of the parent PCI bus, then
zero the BAR and clear the resource list entry for that BAR. This forces
the PCI bus driver to request a valid resource range from the parent bridge
driver when the device driver tries to allocate the BAR. Similarly, if the
initial value of a BAR is a valid range but it is > 4GB and the current OS
only has 32-bit longs, then do a full teardown of the initial value of the
BAR to force a reallocation.
Reviewed by: imp
MFC after: 1 week
of having duplicate versions in each bus attachment.
- Add a DRIVER_MODULE() instance so that the iicbus(4) driver will
actually attach to pcf(4) driver instances.
- Fix compile of envctrl.c.
Pointy hat: jhb (3)
- ad7418(4) uses an sx lock instead of a mtx since the iicbus(4) stuff it
calls can sleep (request_bus()). Also, I expanded the locking slightly
to serialize writes to data stored in the softc.
- Similarly, the icee(4) driver now uses an sx lock instead of a mutex.
I also removed the pointless OPENED flag and flags field from the softc.
- The locking for the ic(4) driver was a bit trickier:
- Add a mutex to the softc to protect softc data.
- The driver uses malloc'd buffers that are the size of the interface
MTU to send and receive packets. Previously, these were allocated
every time the interface was brought up and anytime the MTU was
changed, with various races that could result in memory leaks. I
changed this to be a bit simpler and more like other NIC drivers in
that we allocate buffers during attach for the default MTU size and
only reallocate them on MTU changes. The reallocation procedure
goes to some lengths with various flags to not replace either the
the receive or transmit buffers while the driver is busy receiving
or transmitting a packet.
- Store the device_t of the driver in the softc instead of detours into
new-bus using if_dunit from the ifnet and an even more bizarre detour
to get the softc instead of using if_softc.
- Drop the driver mutex when invoking netisr_dispatch() to pass the
packet up to IP.
- Use if_printf().
- Add an sx lock to the iic(4) driver to serialize open(), close(), read(),
and write and to protect sc_addr and sc_count in the softc.
- Use cdev->si_drv1 instead of using the minor number of the cdev to
lookup the softc via newbus in iic(4).
- Store the device_t in the softc to avoid a similar detour via minor
numbers in iic(4).
- Only add at most one instance of iic(4) and iicsmb(4) to each iicbus(4)
instance, and do it in the child driver.
- Add a mutex to the iicbus(4) softc to synchronize the request/release bus
stuff.
- Use __BUS_ACCESSOR() for IICBUS_ACCESSOR() instead of rolling our own.
- Add a mutex to the iicsmb(4) softc to protect softc state updated in the
interrupt handler.
- Remove Giant from all the smbus methods in iicsmb(4) now that all the
iicbus(4) backend is locked.
- Just grab Giant in the ixp425_iic(4) driver since this driver uses
a shared address/data register window pair to access the actual
I2C registers. None of the other ixp425 drivers lock access to these
shared address/data registers yet and that would need to be done before
this could use any meaningful locking.
- Add locking to the interrupt handler and 'iicbus_reset' methods of the
at91_twi(4) driver.
- Add locking to the pcf(4) driver. Other pcf(4) fixes include:
- Don't needlessly zero the softc.
- Use bus_foo rather than bus_space_foo and remove bus space tag and
handle from softc.
- The lpbb(4) driver just grabs Giant for now. This will be refined later
when ppbus(4) is locked.
- As was done with smbus earlier, move the DRIVER_MODULE() lines to match
the bus driver (either iicbus or iicbb) to the bridge driver into the
bridge drivers.
Tested by: sam (arm/ixp425)
- Add a mutex to the softc to protect the softc and device hardware.
- Use a private timer routine to drive the transmit watchdog timer instead
of using if_watchdog/if_timer.
- If if_alloc() fails during attach, fail the attach with an error rather
than panic'ing.
- Clear RUNNING and OACTIVE only in sncstop().
- Don't mess with IFF_UP.
- Don't leak 'struct ifnet' on detach.
- Setup interrupt handler after ether_ifattach().
- Call ether_ifdetach() rather than if_detach() in the pccard detach
routine.
Tested by: no one despite repeated requests
- let the loop iterate every page (as intended), and not some multiplies
(which led to a fake exhaustion of the ARM_NOCACHE_KVA_SIZE)
- eliminate using MIN(): it compared number of pages vs. address
(ARM_TP_ADDRESS), which was bogus
Reviewed by: cognet, imp
Obtained from: Piotr Ziecik kosmo ! semihalf dot com
MFC after: 3 days
PZERO + 1. The sleeping process at the priority <= PZERO is counted as
blocked, or, as comment states, 'disk wait'. PZERO + 1 works as well,
and does not cause user confusion.
Reported by: sam <samflanker at gmail com>
MFC after: 1 week
commit, calling i386_parsedev(..., X, ...) where X is "ad", "bge", or
any other disk or network device name without a unit number, would
result in dereferencing whatever happened to be on the stack where the
variable "cp" is stored.
Found by: LLVM/Clang Static Checker
to this commit, "env BLOCKSIZE=4X df" prints not only "4X: unknown
blocksize" as expected, but sometimes also "maximum blocksize is 1G"
and "minimum blocksize is 512" depending on what happened to be on
the stack.
Found by: LLVM/Clang Static Checker
watchdog timeout issues and the root cause seems to stem from
silicon bug of controller. Personally I couldn't reproduce it on
RTL8169 controller but it seems it's dependent on usage pattern.
For newer PCIe based controllers I have no TSO complaints but
turning off TSO would be more safe. Users who are sure that
their controller works with TSO can still reenable the TSO with
ifconfig(8).
Reported by: Oliver Lehmann (lehmann at ans-netz dot de), Eugene Butusov (ebutusov at gmail dot com)
11bits. This limits the maximum interface MTU size in TSO case
as upper stack should not generate TCP segments with MSS greater
than the limit. Armed with this information, disable TSO if
interface MTU is greater than the limit.
understand which code paths aren't possible.
This commit eliminates 117 false positive bug reports of the form
"allocate memory; error out if pointer is NULL; use pointer".
environ[0] to be more obvious that environ is not NULL before environ[0]
is tested. Although I believe the previous code worked, this change
improves code maintainability.
Reviewed by: ache
MFC after: 3 days
big endian platforms where time_t is 64bits (ie armeb and sparc64), it will
be a problem.
Use a temporary time_t to work around this.
Submitted by: Matthew Luckie <mjl AT luckie DOT org dot nz>
MFC after: 3 days
assumed to be reviewd by them):
Stir directly from the kernel PRNG, without taking less random pid & time
bytes too (when it is possible).
The difference with OpenBSD code is that they have KERN_ARND sysctl for
that task, while we need to read /dev/random
