Adrian Chadd [Thu, 24 Nov 2011 07:37:19 +0000 (07:37 +0000)]
Now that I've brought up FreeBSD via flash, I've discovered that
the second-last 64k seems to be the default firmware board configuration
area.
Since I have no idea whether uboot uses it or not - and it's prefixed
with an atheros eeprom signature (0xaa55), I figure the safest thing
to do is mark it as read-only.
I've modified my local tplink firmware building program to generate
a board configuration section - which is separate to this partition.
It's located in the 64k _before_ this particular 64k.
The firmware build program from OpenWRT never initialises those
values and the firmware images from tplink also leave it 0x0, so I
don't currently know what the exact, correct details should be.
Adrian Chadd [Thu, 24 Nov 2011 07:32:52 +0000 (07:32 +0000)]
Introduce a new (global, sorry!) option which controls whether
the ar71xx platform code should assume a uboot or redboot environment.
The current code gets very confused (and just crashes) on a uboot
environment, where each attribute=value pair is in a single entry.
Redboot on the other hand stores it as "attribute", "value", "attribute",
"value", ...
This allows the kernel to boot on a TP-LINK TL-WR1043ND from flash,
where the uboot environment gets setup. This didn't show up during a netboot
as "tftpboot" and "go" don't setup the uboot environment variables.
Adrian Chadd [Thu, 24 Nov 2011 04:39:01 +0000 (04:39 +0000)]
Flesh out a geom_map setup, so the kernel can be squeezed _onto_ the device.
The default flash layout gives only 1 megabyte for the kernel, gzipped.
The uboot firmware running on this device only supports gzip, not lzma, so
we actually _do_ have to try and slim the kernel down a bit.
But, since I can't actually do that at the present, I'm opting to:
* extend the kernel from 1mb to 2mb;
* have rootfs fill the rest of that, save 64k;
* eventually I'll hide a 64k config partition at the end, between the
end of rootfs and the ART (radio configuration data.)
The uboot firmware doesn't care about the partition layout. It just
expects the kernel application image to sit at 0xbf020000 (right after
the 128k uboot image.) The uboot header isn't actually read either -
it's "faked" from a "tplink" flash image header. So as long as the
map configuration here matches what is being written out via the
tplink firmware generator, everything is a-ok.
Adrian Chadd [Thu, 24 Nov 2011 04:34:04 +0000 (04:34 +0000)]
Compile in the right bits so the AR9130 WMAC support functions correctly.
A previous commit disabled compiling the AR9130 support in the default
HAL build in the kernel. Since the AR9130 support won't actually function
without AH_SUPPORT_AR9130 (and that abomination needs to be undone at some
point, in order to allow USB 11n NICs to also work), we now have to
explicitly compile it in.
But since the 11n RF backends don't (currently) join the RF linker set,
one has to compile in _an_ RF backend for the HAL to compile.
Adrian Chadd [Thu, 24 Nov 2011 04:23:42 +0000 (04:23 +0000)]
Add a comment documenting where the WMAC hangs off of.
At some point it would be nice to correctly update the bus glue to make
this "correct", including having the DDR flush occur in the right spot
(ie, any AHB interrupt.)
Pyun YongHyeon [Wed, 23 Nov 2011 23:29:18 +0000 (23:29 +0000)]
To save more power, switch to 10/100Mbps link when controller is
put into suspend/shutdown. Old PCI controllers performed that
operation in firmware but for RTL8111C or newer controllers, it's
responsibility of driver. It's not clear whether the firmware of
RTL8111B still downgrades its speed to 10/100Mbps so leave it as it
was.
Pyun YongHyeon [Wed, 23 Nov 2011 22:07:13 +0000 (22:07 +0000)]
Make sure to stop TX MAC before freeing queued TX frames.
For RTL8111DP, check if the TX MAC is active by reading RL_GTXSTART
register. For RTL8402/8168E-VL/8168F/8411, wait until TX queue is
empty.
Marius Strobl [Wed, 23 Nov 2011 22:05:44 +0000 (22:05 +0000)]
Wrap BCM5785 in #ifdef notyet for now. According to yongari@ there are
issues probably needing workarounds in bge(4) when brgphy(4) handles this
PHY. Letting ukphy(4) handle it instead results in a working configuration,
although likely with performance penalties.
Luigi Rizzo [Wed, 23 Nov 2011 09:45:48 +0000 (09:45 +0000)]
fix formatting warning using casts. The numbers involved
are small and these are debug statements, so there is no reason to
obfuscate the format string with PRIsomeKINDofINTEGER
Pyun YongHyeon [Wed, 23 Nov 2011 02:08:05 +0000 (02:08 +0000)]
Disable accepting frames in re_stop() to put RX MAC into idle state.
Because there is no reliable way to know whether RX MAC is in
stopped state, rejecting all frames would be the only way to
minimize possible races.
Otherwise it's possible to receive frames while stop command
execution is in progress and controller can DMA the frame to freed
RX buffer during that period.
This was observed on recent PCIe controllers(i.e. RTL8111F).
While this change may not be required on old controllers it
wouldn't make negative effects on old controllers. One side effect
of this change is disabling receive so driver reprograms RL_RXCFG
to receive WOL frames when it is put into suspend or shutdown.
This should address occasional 'memory modified free' errors seen
on recent RealTek controllers.
Pyun YongHyeon [Tue, 22 Nov 2011 23:27:59 +0000 (23:27 +0000)]
Perform media change after setting IFF_DRV_RUNNING flag. Without it,
driver would ignore the first link state update if controller
already established a link such that it would have to take
additional link state handling in re_tick().
Pyun YongHyeon [Tue, 22 Nov 2011 23:19:49 +0000 (23:19 +0000)]
Writing access to RL_CFG5 register also requires EEPROM write
access.
While I'm here, enable WOL through magic packet but disable waking
up system via unicast, multicast and broadcast frames. Otherwise,
multicast or unicast frame(e.g. ICMP echo request) can wake up
system which is not probably wanted behavior on most environments.
This was not known as problem because RL_CFG5 register access had
not effect until this change.
The capability to wake up system with unicast/multicast frames
are still set in driver, default off, so users who need that
feature can still activate it with ifconfig(8).
Marius Strobl [Tue, 22 Nov 2011 21:28:20 +0000 (21:28 +0000)]
- There's no need to overwrite the default device method with the default
one. Interestingly, these are actually the default for quite some time
(bus_generic_driver_added(9) since r52045 and bus_generic_print_child(9)
since r52045) but even recently added device drivers do this unnecessarily.
Discussed with: jhb, marcel
- While at it, use DEVMETHOD_END.
Discussed with: jhb
- Also while at it, use __FBSDID.
Pyun YongHyeon [Tue, 22 Nov 2011 21:22:06 +0000 (21:22 +0000)]
For IP1001 PHY, do not set multi-port device(MASTER). Ideally this
bit should not affect link establishment process of auto-negotiation
if manual configuration is not used, which is true in auto-negotiation.
However it seems setting this bit interfere with IP1001 PHY's
down-shifting feature such that establishing a 10/100Mbps link failed
when 1000baseT link is not available during auto-negotiation process.
Tested by: Andrey Smagin <samspeed <> mail dot ru >
Pyun YongHyeon [Tue, 22 Nov 2011 20:57:06 +0000 (20:57 +0000)]
Announce flow control capability to underlying PHY driver.
Pause timer value is initialized to 0xFFFF. Controller allows just
4 different TX pause thresholds. The lowest possible threshold
value looks too aggressive so use next available threshold value.
Mikolaj Golub [Tue, 22 Nov 2011 20:45:11 +0000 (20:45 +0000)]
Retire linprocfs_doargv(). Instead use new functions, proc_getargv()
and proc_getenvv(), which were implemented using linprocfs_doargv() as
a reference.
Pyun YongHyeon [Tue, 22 Nov 2011 20:45:09 +0000 (20:45 +0000)]
Rework link establishment and link state detection logic.
- Remove MIIBUS statchg callback and program VGE_DIAGCTL before
initiating link establishment. Previously driver used to
program VGE_DIAGCTL after getting a link in statchg callback.
It seems the VGE_DIAGCTL register works like a kind of MII
register such that it requires setting a 'to be' mode in advance
rather than relying on resolved speed/duplex of established link.
This means the statchg callback is not needed in driver. In
addition, if there was no link at the time of media change, this
was not called at all.
- Introduce vge_ifmedia_upd_locked() to change current media to
configured one. Actual media change is performed only after PHY
reset and VGE_DIAGCTL setup.
- In WOL configuration, make sure to clear forced mode such that
controller can rely on auto-negotiation.
- Unlike most other drivers that use miibus(4), vge(4) used
controller's auto-polling feature for link state tracking via
interrupt. This came from controller's inefficient mechanism to
access MII registers. On link state change interrupt, vge(4)
used to get current link state with series of MII register
accesses. Because vge(4) already enabled auto polling, read PHY
status register to resolved speed/duplex/flow control parameters.
vge(4) still does not drive MII_TICK to reduce number of MII
register accesses which in turn means the driver does not know the
status of auto-negotiation. This was a one of long standing
issue of vge(4). Probably driver may be able to implement a timer
that keeps track of auto-negotiation state and restart
auto-negotiation when driver couldn't establish a link within a
specified period. However the controller does not provide a
reliable way to detect auto-negotiation failure so I'm not sure
whether it's worth to implement it in driver.
Alternatively driver can completely disable MII auto-polling and
let miibus(4) poll link state by driving MII_TICK. This may reduce
unnecessary overhead of stopping/restarting MII auto-polling of
controller. Unfortunately it was known that some variants of
controller does not work correctly if MII auto-polling is disabled.
Marius Strobl [Tue, 22 Nov 2011 19:31:43 +0000 (19:31 +0000)]
- Add a DEVMETHOD_END alias for KOBJMETHOD_END so that along with 'driver_t'
and DEVMETHOD() we can fully hide the explicit mention of kobj(9) from
device drivers.
- Update the example in driver.9 to use DEVMETHOD_END.
Robert Millan [Tue, 22 Nov 2011 18:53:54 +0000 (18:53 +0000)]
Define __FreeBSD_kernel__ macro in sys/param.h.
__FreeBSD_kernel__ indicates that this system uses the kernel of FreeBSD,
which by definition is always true on FreeBSD. This macro is also defined
on other systems that use the kernel of FreeBSD, such as GNU/kFreeBSD.
It is tempting to use this macro in userland code when we want to enable
kernel-specific routines, and in fact it's fine to do this in code that
is part of FreeBSD itself. However, be aware that as presence of this
macro is still not widespread (e.g. older FreeBSD versions, 3rd party
compilers, etc), it is STRONGLY DISCOURAGED to check for this macro in
external applications without also checking for __FreeBSD__ as an
alternative.
Rick Macklem [Tue, 22 Nov 2011 00:35:30 +0000 (00:35 +0000)]
This patch enables the new/default NFS server's use of shared
vnode locking for read, readdir, readlink, getattr and access.
It is hoped that this will improve server performance for these
operations, since they will no longer be serialized for a given
file/vnode.
Adrian Chadd [Mon, 21 Nov 2011 22:57:28 +0000 (22:57 +0000)]
Use the correct lock when calling msleep().
This fixes panics that users have been seeing when operating in station mode,
where the interface undergoes a lot more resets then in hostap mode (ie whilst
doing channel scanning.)
Reported by: arundel, wblock@wonkity.com
Sponsored by: Hobnob, Inc.
Bernhard Schmidt [Mon, 21 Nov 2011 22:19:12 +0000 (22:19 +0000)]
The DC calibration result obtained during initialization can't be
passed over to the runtime firmware on 6050 devices. Instead let
the runtime firmware do the calibration itself. This fixes support
for the 6050 series devices.
Martin Matuska [Mon, 21 Nov 2011 20:33:18 +0000 (20:33 +0000)]
Revert back to revision 227649 because of license uncertainity.
Manual pages from OpenSolaris svn_134 are still properly CDDL licensed
but I have been informed that the parts from s11ex are uncertain even
if they contain a CDDL header.
Xin LI [Mon, 21 Nov 2011 20:26:22 +0000 (20:26 +0000)]
Improve the way to calculate available pages in tmpfs:
- Don't deduct wired pages from total usable counts because it does not
make any sense. To make things worse, on systems where swap size is
smaller than physical memory and use a lot of wired pages (e.g. ZFS),
tmpfs can suddenly have free space of 0 because of this;
- Count cached pages as available; [1]
- Don't count inactive pages as available, technically we could but that
might be too aggressive; [1]
Simplify the libpam build by removing the shared modules' dependency
on the shared library. The modules are loaded by the library, so we
know it'll be there when we need it.
Martin Matuska [Mon, 21 Nov 2011 15:26:20 +0000 (15:26 +0000)]
Update zfs(8) and zpool(8) manpages from CDDL-licensed sources [1].
Improved alignment for a maximum width of 80 characters.
Mark unsupported parts as such.
Gleb Smirnoff [Mon, 21 Nov 2011 14:10:13 +0000 (14:10 +0000)]
Historically in_control() did not check sockaddrs supplied with
structs ifreq/in_aliasreq and there've been several panics due
to that problem. All these panics were fixed just a couple of
lines above the panicing code.
Take a more general approach: sanity check sockaddrs supplied
with SIOCAIFADDR and SIOCSIF*ADDR at the beggining of the
function and drop all checks below.
One check is now disabled due to strange code in ifconfig(8)
that I've removed recently. I'm going to enable it with next
__FreeBSD_version bump.
Historically in_ifinit() was able to recover from an error
and restore old address. Nowadays this feature isn't working
for all error cases, but for some of them. I suppose no software
relies on this behavior, so I'd like to remove it, since this
simplifies code a lot.
Also, move if_scrub() earlier in the in_ifinit(). It is more
correct to wipe routes before removing address from local
address list, and interface address list.
Lawrence Stewart [Mon, 21 Nov 2011 13:34:29 +0000 (13:34 +0000)]
- Add Pulse-Per-Second timestamping using raw ffcounter and corresponding
ffclock time in seconds.
- Add IOCTL to retrieve ffclock timestamps from userland.
Committed on behalf of Julien Ridoux and Darryl Veitch from the University of
Melbourne, Australia, as part of the FreeBSD Foundation funded "Feed-Forward
Clock Synchronization Algorithms" project.
For more information, see http://www.synclab.org/radclock/
Submitted by: Julien Ridoux (jridoux at unimelb edu au)
Gleb Smirnoff [Mon, 21 Nov 2011 12:07:18 +0000 (12:07 +0000)]
- Reduce severity for all ARP events, that can be triggered from remote
machine to LOG_NOTICE. Exception left to "using my IP address".
- Fix multicast ARP warning: add newline and also log the bad MAC address.