return to user space w/o waiting for I/O to complete.
I tried to get several folks who know this code better than me to review it
with no luck. I *do* know that w/o this code, using the SCSI target driver
panics in userret (if it doesn't panic in knote first).
if a process's uid or gid has changed, but the /proc/<PID> directory
itself was also set to mode 0. Assuming this doesn't open any
security holes, open access to the /proc/<PID> directory for users
other than root to read or search the directory.
Reviewed by: des (back in February)
MFC after: 3 weeks
Since tags are kept while packet resides in kernelspace, it's possible to
use other kernel facilities (like netgraph nodes) for altering those tags.
Submitted by: Andrey Elsukov <bu7cher at yandex dot ru>
Submitted by: Vadim Goncharov <vadimnuclight at tpu dot ru>
Approved by: glebius (mentor)
Idea from: OpenBSD PF
MFC after: 1 month
SRC_BASE package (src/[A-Z]*) as well as SRC_SYS (src/sys/*). This
allows users who only install the kernel source code to use the
modern "make buildkernel" approach.
Discussed with: re (scottl, kensmith)
MFC after: 3 days
EHCI spec for linking in new qTDs into an asynchronous QH. This
requires that there is a qTD marked as not active and not halted
at the start of the QH's list, and the hardware will know to re-fetch
the qTD on each pass rather than just looking at the overlay qTD:
"The host controller must be able to advance the queue from the
Fetch QH state in order to avoid all hardware/software race
conditions. This simple mechanism allows software to simply link
qTDs to the queue head and activate them, then the host controller
will always find them if/when they are reachable."
This is achieved by keeping an "inactivesqtd" entry on the QH list,
and re-using it each time as the start of the next transfer, and
allocating a new qTD to become the next inactivesqtd. Then a new
transfer can be activated by just setting its "active" flag, which
avoids all the previous messing with overlay qTD state in
ehci_set_qh_qtd().
more appropriate value and is also the default set by the kernel. I
could not find a justification of why rc.conf began overriding it back
in 1998.
This dramatically cuts NFS traffic on e.g. a busy system with NFS root.
Reviewed by: mohans
MFC After: 2 weeks
mimicing the NFS reference implementation.
NFS over TCP does not need fast retransmit timeouts, since network loss
and congestion are managed by the transport (TCP), unlike with NFS over
UDP. A long timeout prevents the unnecessary retransmission of non-
idempotent NFS requests.
Reviewed by: mohans, silby, rees?
Sponsored by: Network Appliance, Incorporated
the estimator to be more easily tuned and maintained.
There should be no functional change except there is now a lower limit
on the retransmit timeout to prevent the client from retransmitting
faster than the server's disks can fill requests, and an upper limit
to prevent the estimator from taking to long to retransmit during a
server outage.
Reviewed by: mohan, kris, silby
Sponsored by: Network Appliance, Incorporated
before starting exploring (4 seconds), and extend the wait period
if new USB buses are attached while waiting.
This works around a problem seen when there is more than one EHCI
controller in the system and you kldload usb.ko after the system
has booted. The problem is that usb.ko contains 3 separate PCI
drivers which get initialised one by one (uhci, ohci, ehci), and
when each driver is initialised, all PCI buses are re-probed after
just the addition of that driver. This means that there can be a
significant delay between the attaching of a companion controller
and the subsequent EHCI attach, so it is possible for the companion
controller's USB 1.x bus to be scanned before the EHCI driver gets
a chance to check if there is really a USB 2.x device connected.
- Rename REG_DL to REG_DLL and REG_DLH.
- Always treat DLL and DLH as two separate 8-bit registers instead of one
16-bit register.
Additionally, remove the probe for the high 4 bits of IER being 0 and don't
assume we can always read/write 0 to/from those bits.
These changes allow uart(4) to drive the UARTs on the Intel XScale PXA255.
Reviewed by: marcel
parameter optional.
- Add Read_Node_List command which prints a list of available HCI nodes,
their Netgraph IDs and connected hooks
Reviewed by: emax
Approved by: emax
MFC after: 1 week
- Skip PnP devices as some wedge when trying to probe them as C-NET(98)S.
This fix makes le(4) actually work with the C-NET(98)S.
Reviewed by: marius
Tested by: Watanabe Kazuhiro < CQG00620 at nifty dot ne dot jp >
Checking if the queues are empty is not enough for the crypto_proc thread
(it is enough for the crypto_ret_thread), because drivers can be marked
as blocked. In a situation where we have operations related to different
crypto drivers in the queue, it is possible that one driver is marked as
blocked. In this case, the queue will not be empty and we won't wakeup
the crypto_proc thread to execute operations for the others drivers.
Simply setting a global variable to 1 when we goes to sleep and setting
it back to 0 when we wake up is sufficient. The variable is protected
with the queue lock.
Before the change if the thread was working on symmetric operation, we
would send unnecessary wakeup after adding asymmetric operation (when
asym queue was empty) and vice versa.
twice if we call crypto_kinvoke() from crypto_proc thread.
This change also removes unprotected access to cc_kqblocked field
(CRYPTO_Q_LOCK() should be used for protection).
where crypto_invoke() returns ERESTART and before we set cc_qblocked to 1,
crypto_unblock() is called and sets it to 0. This way we mark device as
blocked forever.
Fix it by not setting cc_qblocked in the fast path and by protecting
crypto_invoke() in the crypto_proc thread with CRYPTO_Q_LOCK().
This won't slow things down, because there is no contention - we have
only one crypto thread. Actually it can be slightly faster, because we
save two atomic ops per crypto request.
The fast code path remains lock-less.
