written by Stuart Walsh and Duncan Barclay (with some kibbitzing by
me). I'm checking it in on Stuart's behalf.
The BCM4401 is built into several x86 laptop and desktop systems. For the
moment, I have only enabled it in the x86 kernel config because although
it's a PCI device, I haven't heard of any standalone NICs that use it. If
somebody knows of one, we can easily add it to the other arches.
This driver uses register/structure data gleaned from the Linux
driver released by Broadcom, but does not contain any of the code
from the Linux driver itself. It uses busdma.
latter is a kernel option for IA64_ID_PAGE_SHIFT, which in turn
determines IA64_ID_PAGE_MASK and IA64_ID_PAGE_SIZE.
The constants are used instead of the literal hardcoding (in its
various forms) of the size of the direct mappings created in region
6 and 7. The default and probably only workable size is still 256M,
but for kicks we use 128M for LINT.
specified directory is not found in the mount list. Before the
MNT_BYFSID changes, unmount(2) used to return ENOENT for a nonexistent
path and EINVAL for a non-mountpoint, but we can no longer distinguish
between these cases. Of the two error codes, EINVAL was more likely
to occur in practice, and it was the only one of the two that was
documented.
Update the manual page to match the current behaviour.
Suggested by: tjr
Reviewed by: tjr
atomically extracts and holds the physical page that is associated with the
given pmap and virtual address. Such a function is needed to make the
memory mapping optimizations used by, for example, pipes and raw disk I/O
MP-safe.
Reviewed by: tegge
rl(4) driver and put it in a new re(4) driver. The re(4) driver shares
the if_rlreg.h file with rl(4) but is a separate module. (Ultimately
I may change this. For now, it's convenient.)
rl(4) has been modified so that it will never attach to an 8139C+
chip, leaving it to re(4) instead. Only re(4) has the PCI IDs to
match the 8169/8169S/8110S gigE chips. if_re.c contains the same
basic code that was originally bolted onto if_rl.c, with the
following updates:
- Added support for jumbo frames. Currently, there seems to be
a limit of approximately 6200 bytes for jumbo frames on transmit.
(This was determined via experimentation.) The 8169S/8110S chips
apparently are limited to 7.5K frames on transmit. This may require
some more work, though the framework to handle jumbo frames on RX
is in place: the re_rxeof() routine will gather up frames than span
multiple 2K clusters into a single mbuf list.
- Fixed bug in re_txeof(): if we reap some of the TX buffers,
but there are still some pending, re-arm the timer before exiting
re_txeof() so that another timeout interrupt will be generated, just
in case re_start() doesn't do it for us.
- Handle the 'link state changed' interrupt
- Fix a detach bug. If re(4) is loaded as a module, and you do
tcpdump -i re0, then you do 'kldunload if_re,' the system will
panic after a few seconds. This happens because ether_ifdetach()
ends up calling the BPF detach code, which notices the interface
is in promiscuous mode and tries to switch promisc mode off while
detaching the BPF listner. This ultimately results in a call
to re_ioctl() (due to SIOCSIFFLAGS), which in turn calls re_init()
to handle the IFF_PROMISC flag change. Unfortunately, calling re_init()
here turns the chip back on and restarts the 1-second timeout loop
that drives re_tick(). By the time the timeout fires, if_re.ko
has been unloaded, which results in a call to invalid code and
blows up the system.
To fix this, I cleared the IFF_UP flag before calling ether_ifdetach(),
which stops the ioctl routine from trying to reset the chip.
- Modified comments in re_rxeof() relating to the difference in
RX descriptor status bit layout between the 8139C+ and the gigE
chips. The layout is different because the frame length field
was expanded from 12 bits to 13, and they got rid of one of the
status bits to make room.
- Add diagnostic code (re_diag()) to test for the case where a user
has installed a broken 32-bit 8169 PCI NIC in a 64-bit slot. Some
NICs have the REQ64# and ACK64# lines connected even though the
board is 32-bit only (in this case, they should be pulled high).
This fools the chip into doing 64-bit DMA transfers even though
there is no 64-bit data path. To detect this, re_diag() puts the
chip into digital loopback mode and sets the receiver to promiscuous
mode, then initiates a single 64-byte packet transmission. The
frame is echoed back to the host, and if the frame contents are
intact, we know DMA is working correctly, otherwise we complain
loudly on the console and abort the device attach. (At the moment,
I don't know of any way to work around the problem other than
physically modifying the board, so until/unless I can think of a
software workaround, this will have do to.)
- Created re(4) man page
- Modified rlphy.c to allow re(4) to attach as well as rl(4).
Note that this code works for the sample 8169/Marvell 88E1000 NIC
that I have, but probably won't work for the 8169S/8110S chips.
RealTek has sent me some sample NICs, but they haven't arrived yet.
I will probably need to add an rlgphy driver to handle the on-board
PHY in the 8169S/8110S (it needs special DSP initialization).
ia64_count_cpus() and ia64_probe_sapics() called a single function
to do the the actual work. The difference in behaviour was handled
in that function and was further complicated by adding bootverbose
related code. As such, even the simplest of changes was hard to
comprehend.
Untangling has been done by increasing code duplication and using
a more naive style of coding. FWIW, the object file is slightly
smaller than before, so things aren't as bad as it may seem.
Triggered by: a simple fix on the P4 branch that never got merged.
from the SAB82532 and the Z8530 hardware drivers by introducing
uart_cpu_busaddr(). The assumption is not true on pc98 where
bus_space_handle_t is a pointer to a structure.
The uart_cpu_busaddr() function will return the bus address
corresponding the tag and handle given to it by the BAS.
WARNING: the intend of the function is STRICTLY to allow hardware
drivers to determine which logical channel they control and is NOT
to be used for actual I/O. It is therefore EXPLICITLY allowed that
uart_cpu_busaddr() returns only the lower 8 bits of the address
and garbage in all other bits. No mistakes...
Quick fix for calling DELAY() for ddb input in some (atkbd-based)
console drivers. ddb must not use any normal locks, but DELAY()
normally calls getit() which needs clock_lock. One problem with using
normal locks in ddb is that deadlock is possible, but deadlock on
clock_lock is unlikely becaluse clock_lock is bogusly recursive,
apparently just to hide the problem of ddb using it. The i8254 clock
hardware has mostly write-only registers so it is important for it to
use a lock that gives exclusive access. (atkbd hardware is also
unfriendly to reentrant software but that problem is more local and
already solved.) I mostly saw the symptoms of the bug caused by
unlocking in getit() running cpu_unpend(). cpu_unpend() should not
be called while in ddb and Debugger() calls for failing assertions
about this caused a breakpoint within ddb.
ddb must also not call getit() because ddb may be being used to step
through clock initialization code that has stopped or otherwise mangled
the clock. If the clock is stopped, then getit() always returns the
same value and DELAY() takes forever if it trusts getit().
The quick fix is implement DELAY(n) as (n * timer_freq / 1000000)
inb(0x84)'s if ddb is active.
machdep.c:
Don't permit recursion on clock_lock.
kdb_trap(). Stopping the other CPUs acts like locking them out, but
it wasn't done early enough or held long enough to prevent concurrent
accesses to shared data. In particular, the saved regs could be
clobbered.
with 64-bit longs again. This was fixed in rev.1.42 but the fix
rotted non-fatally in rev.1.105 and fatally in rev.1.137.
Many more non-egregrious casts are strictly required for conversions
from semi-opaque types to pointers, but we avoid most of them by using
types that are almost certain to be compatible with uintptr_t for
representing pointers (e.g., vm_offset_t). Here we don't really want
the u_longs, but we have them because a.out.h and its support code
doesn't use typedefs (it uses unsigned in V7 and unsigned long in
FreeBSD) and is too obsolete to fix now.
FIDs to be 128-bits wide and adds support for realms.
Add a new CODA_COMPAT_5 option, which requests support for the old
Coda 5.x interface instead of the new one.
Create a new coda5.ko module that supports the 5.x interface, and make
the existing coda.ko module use the new 6.x interface. These modules
cannot both be loaded at the same time.
Obtained from: Jan Harkes & the coda-6.0.2 distribution,
NetBSD (drochner) (CODA_COMPAT_5 option).
building a module. Inclusion of option files (opt_ddb.h in this
case) is not possible for modules. The inclusion of opt_ddb.h
in this header is questionable.
(ns8250 copied and s/ns8250/i8251/g), but there for linkage purposes.
Real code to follow, once I get past some boot issues on my pc98 boxes
with recent current.
of what uart(4) is and/or is not see the initial commit log of one
of the files in sys/dev/uart (or see share/man/man4/uart.4).
Note that currently pc98 shares the MD file with i386. This needs
to change when pc98 support is fleshed-out to properly support the
various UARTs. A good example is sparc64 in this respect.
We build uart(4) as a module on all platforms. This may break
the ppc port. That depends on whether they do actually build
modules.
To use uart(4) on alpha, one must use the NO_SIO option.
It improves on sio(4) in the following areas:
o Fully newbusified to allow for memory mapped I/O. This is a must
for ia64 and sparc64,
o Machine dependent code to take full advantage of machine and firm-
ware specific ways to define serial consoles and/or debug ports.
o Hardware abstraction layer to allow the driver to be used with
various UARTs, such as the well-known ns8250 family of UARTs, the
Siemens sab82532 or the Zilog Z8530. This is especially important
for pc98 and sparc64 where it's common to have different UARTs,
o The notion of system devices to unkludge low-level consoles and
remote gdb ports and provides the mechanics necessary to support
the keyboard on sparc64 (which is UART based).
o The notion of a kernel interface so that a UART can be tied to
something other than the well-known TTY interface. This is needed
on sparc64 to present the user with a device and ioctl handling
suitable for a keyboard, but also allows us to cleanly hide an
UART when used as a debug port.
Following is a list of features and bugs/flaws specific to the ns8250
family of UARTs as compared to their support in sio(4):
o The uart(4) driver determines the FIFO size and automaticly takes
advantages of larger FIFOs and/or additional features. Note that
since I don't have sufficient access to 16[679]5x UARTs, hardware
flow control has not been enabled. This is almost trivial to do,
provided one can test. The downside of this is that broken UARTs
are more likely to not work correctly with uart(4). The need for
tunables or knobs may be large enough to warrant their creation.
o The uart(4) driver does not share the same bumpy history as sio(4)
and will therefore not provide the necessary hooks, tweaks, quirks
or work-arounds to deal with once common hardware. To that extend,
uart(4) supports a subset of the UARTs that sio(4) supports. The
question before us is whether the subset is sufficient for current
hardware.
o There is no support for multiport UARTs in uart(4). The decision
behind this is that uart(4) deals with one EIA RS232-C interface.
Packaging of multiple interfaces in a single chip or on a single
expansion board is beyond the scope of uart(4) and is now mostly
left for puc(4) to deal with. Lack of hardware made it impossible
to actually implement such a dependency other than is present for
the dual channel SAB82532 and Z8350 SCCs.
The current list of missing features is:
o No configuration capabilities. A set of tunables and sysctls is
being worked out. There are likely not going to be any or much
compile-time knobs. Such configuration does not fit well with
current hardware.
o No support for the PPS API. This is partly dependent on the
ability to configure uart(4) and partly dependent on having
sufficient information to implement it properly.
As usual, the manpage is present but lacks the attention the
software has gotten.
o Introduce PUC_PORT_TYPE_UART so that we can attach to uart(4),
o Introduce port sub-types (eg PUC_PORT_UART_NS8250, PUC_PORT_UART_Z8530)
to handle different hardware and determine resource sizes.
o Introduce two new IVARs: PUC_IVAR_SUBTYPE and PUC_IVAR_REGSHFT. Both
are used by uart(4) to get sufficient information to talk to the HW.
o Introduce PUC_FLAGS_ALTRES to tell puc(4) to try memory mapped I/O
if I/O port space cannot be allocated, or vice versa.
o Have ports of type PUC_PORT_TYPE_COM attach to uart(1) if attaching
to sio(4) fails (due to not having the sio driver).
o Put struct puc_device_description in struct puc_softc instead of
having a pointer to a device description in the softc. This allows
us to create device descriptions on the fly without having to use
malloc() or otherwise have them staticly defined.
o Move puc_find_description() from puc.c to puc_pci.c as it's specific
to PCI.
o Add EBUS and SBUS frontends for use on sparc64. Note that the P in
puc stands for PCI, so we kinda mess things up here. It's too soon
to worry about it though. We'll know what to do about it in time.
NOTE: This commit changes the behaviour of puc(4) to not quieten the
device probe and attach for child devices. The uart(4) driver provides
additional device description that is valuable to have.
we actually use. Originally, the code reserved 0x8000 to 0x80ff inclusive
which on my hardware conflicts with the acpi timer. This broke the amdpm
driver since it was actually given ports 0x800c to 0x810b (which should
not have happened, IMHO).
This also allows us to considerably simplify the handling of the nForce
smb driver, removing the need for a separate nfpm driver. With this, SMB
accesses appear to work on my Tyan Tiger MP board. Your mileage may vary.
In particular, the nForce changes have not been tested.
we can switch to 64M-sized identity mappings and not having to map the
first 64M. This is especially important because the first 1M contains
the VGA frame buffer and is otherwise a legacy memory range. Best to
make as little assumptions about it as possible. Switching to 64M-sized
mappings is important to avoid creating overlapping translations, which
have the side-effect of triggering machine checks. This is currently
what's preventing us to boot on an Intel Tiger 4.
Note that since we currently use 256M-sized identity mappings, we
would reduce the size of the mappings and consequently increase the
TLB pressure. The performance implications of this are minimal if
measurable at all because identify mappings are not our primary
means for memory management.
Also note that there's no guarantee that physical memory exists at
64M. Then again, we didn't had the guarantee when we were loading at
5M. We'll deal with this when it's a problem.
Discussed with: arun@
Special thanks to Pavlin Radoslavov <pavlin@icir.org> for testing and
fixing numerous problems.
Sponsored by: FreeBSD Foundation
Reviewed by: Pavlin Radoslavov <pavlin@icir.org>
and/or INTR_FAST. This belongs elsehwere and perhaps under bootverbose;
I'm committing it for now as it's uesful to know which drivers have
been converted and which have not.
we return to kernel or userland. This triggered a panic in a KSE
application when TDF_USTATCLOCK was set in the case userland was
interrupted, but we never called ast() on our way out. As such,
we called ast() at some other time. Unfortunately, TDF_USTATCLOCK
handling assumes running in the interrupt thread. This was not
the case anymore.
To avoid making the same mistake later, interrupt() now returns
to its caller whether we interrupted userland or not. This avoids
that we have to duplicate the check in assembly, where it's bound
to fall off the scope. Now we simply check the return value and
call ast() if appropriate.
Run into this: davidxu
to protect the vlan state in each ifnet (e.g. vlan count). The latter is
probably better handled through an ifnet-centric means but since changes
are infrequent shouldn't matter for now.
Sponsored by: FreeBSD Foundation
For the floppy driver, use fdcontrol to manipulate density selection.
For the CD drivers, the 'a' and 'c' suffix is without actual effect and
any applications insisting on it can be satisfied with a symlink:
ln -s /dev/cd0 /dev/cd0a
Ongoing discussion may result in these pieces of code being removed before
the 5-stable branch as opposed to after.
such a card is ejected, we'd panic. Instead, just ignore it.
I should also add a sanity check in the FUNCID code as well, but this
isn't wrong since the check is cheap and happens infrequently.
into targreadfilt(). Unlock around calls to notify_user(). If an application
is sending CCBs while the endpoint is shutting down, this may result in
incomplete disable. A more complete solution will come with a "dying" flag.
Submitted by: simokawa
a correctable DMA error. Failing to do so can cause the error interrupt
to be triggered over and over again.
- Clean up the comments for UEAFSR_* constants, fix a typo (UEAFSR_BLK is
(1 << 23), not (1 << 22)), and add two more. Also, add similar constants
for the CE AFSR bits.
We can't update the device description in attach (why not ?), so
we device_print() what we find.
Conditionalize the short cable fix on this being older than rev 16A.
Call device_printf() when we apply short cable fix.
Include interrupt hold-off setting for rev 16+ under "#ifdef notyet"
The device_printf()'s will go under bootverbose once the various
issues have settled a bit.
out of cdregister() and daregister(), which are run from interrupt context.
The sysctl code does blocking mallocs (M_WAITOK), which causes problems
if malloc(9) actually needs to sleep.
The eventual fix for this issue will involve moving the CAM probe process
inside a kernel thread. For now, though, I have fixed the issue by moving
dynamic sysctl variable creation for these two drivers to a task queue
running in a kernel thread.
The existing task queues (taskqueue_swi and taskqueue_swi_giant) run in
software interrupt handlers, which wouldn't fix the problem at hand. So I
have created a new task queue, taskqueue_thread, that runs inside a kernel
thread. (It also runs outside of Giant -- clients must explicitly acquire
and release Giant in their taskqueue functions.)
scsi_cd.c: Remove sysctl variable creation code from cdregister(), and
move it to a new function, cdsysctlinit(). Queue
cdsysctlinit() to the taskqueue_thread taskqueue once we
have fully registered the cd(4) driver instance.
scsi_da.c: Remove sysctl variable creation code from daregister(), and
move it to move it to a new function, dasysctlinit().
Queue dasysctlinit() to the taskqueue_thread taskqueue once
we have fully registered the da(4) instance.
taskqueue.h: Declare the new taskqueue_thread taskqueue, update some
comments.
subr_taskqueue.c:
Create the new kernel thread taskqueue. This taskqueue
runs outside of Giant, so any functions queued to it would
need to explicitly acquire/release Giant if they need it.
cd.4: Update the cd(4) man page to talk about the minimum command
size sysctl/loader tunable. Also note that the changer
variables are available as loader tunables as well.
da.4: Update the da(4) man page to cover the retry_count,
default_timeout and minimum_cmd_size sysctl variables/loader
tunables. Remove references to /dev/r???, they aren't used
any longer.
cd.9: Update the cd(9) man page to describe the CD_Q_10_BYTE_ONLY
quirk.
taskqueue.9: Update the taskqueue(9) man page to describe the new thread
task queue, and the taskqueue_swi_giant queue.
MFC after: 3 days
in a list head instead of a pointer to the first element at the time of
the first call. These lists are subject to change, and getdirtybuf()
would refetch from the wrong list in some cases.
Spottedy by: tegge
Pointy hat to: me
address of the device identified by its phandle_t by traversing OFW's
device tree. The space and address returned by this function can
subsequently be passed to sparc64_fake_bustag() to construct a valid
tag and handle for use by the newbus I/O functions.
Use of this function is expected to be limited to pre-newbus access to
devices, such as consoles and keyboards.
Partially obtained from: tmm
Reviewed by: jake, jmg, tmm
SBus testing made possible by: jake
Tested with: LINT
and replace it with the more intuitive name PCIR_BARS.
- Add a PCIR_BAR(x) macro that returns the config space register offset of
the 32-bit BAR x.
MFC after: 3 days
This replaces the current ioctl processing with a direct call path
from geom_dev() where the ioctl arrives (from SPECFS) to any directly
connected GEOM class.
The inverse of the above is no longer supported. This is the
situation were you have one or more intervening GEOM classes, for
instance a BSDlabel on top of a MBR or PC98. If you want to issue
MBR or PC98 specific ioctls, you will need to issue them on a MBR
or PC98 providers.
This paves the way for inviting CD's, FD's and other special cases
inside GEOM.
it in the last chunk (phys_avail block). The last chunk very often is
not larger than one or two pages, resulting in a msgbuf that's too
small to hold a complete verbose boot.
Note that pmap_steal_memory() will bzero the memory it "allocates".
Consequently, ia64 will never preserve previous msgbufs. This is not
a noticable difference in practice. If the msgbuf could be reused,
it was invariably too small to have anything preserved anyway.
Changes from the original implementation:
- Fragmentation is handled by the function m_fragment, which can
be called from whereever fragmentation is needed. Note that this
function is wrapped in #ifdef MBUF_STRESS_TEST to discourage non-testing
use.
- m_fragment works slightly differently from the old fragmentation
code in that it allocates a seperate mbuf cluster for each fragment.
This defeats dma_map_load_mbuf/buffer's feature of coalescing adjacent
fragments. While that is a nice feature in practice, it nerfed the
usefulness of mbuf_stress_test.
- Add two modes of random fragmentation. Chains with fragments all of
the same random length and chains with fragments that are each uniquely
random in length may now be requested.
o add locking
o strip irrelevant spl's
o split malloc types to better account for memory use
o remove unused IPSEC_NONBLOCK_ACQUIRE code
o remove dead code
Sponsored by: FreeBSD Foundation
NB: There is a known LOR on the forwarding path; this needs to be resolved
together with a similar issue in the bridge. For the moment it is
believed to be benign.
Sponsored by: FreeBSD Fondation
o remove irrlevant spl
Notes:
1. We don't lock domain list traversals as this is safe until we start
removing domains.
2. The calculation of max_datalen in net_init_domain appears safe as
noone depends on max_hdr and max_datalen having consistent values.
3. Giant is still held for fast and slow timeouts; this must stay until
each timeout routine is properly locked (coming soon).
Sponsored by: FreeBSD Fondation
bail out if the buffer is not already present.
- The buffer returned by incore() is not locked and should not be sent to
brelse(). Use getblk() with the new GB_NOCREAT flag to preserve the
desired semantics.
caller to acquire it. This permits drain_output() to be done atomically
with other operations as well as reducing the number of lock operations.
- Assert that the proper locks are held in drain_output().
- Change getdirtybuf() to accept a mutex as an argument. This mutex is used
to protect the vnode's buf list and the BKGRDWAIT flag. This lock is
dropped when we successfully acquire a buffer and held on return
otherwise. These semantics reduce the number of cumbersome cases in
calling code.
- Pass the mtx from getdirtybuf() into interlocked_sleep() and allow this
mutex to be used as the interlock argument to BUF_LOCK() in the LOCKBUF
case of interlocked_sleep().
- Change the return value of getdirtybuf() to be the resulting locked buffer
or NULL otherwise. This is for callers who pass in a list head that
requires a lock. It is necessary since the lock that protects the list
head must be dropped in getdirtybuf() so that we don't have a lock order
reversal with the buf queues lock in bremfree().
- Adjust all callers of getdirtybuf() to match the new semantics.
- Add a comment in indir_trunc() that points at unlocked access to a buf.
This may also be one of the last instances of incore() in the tree.