yongari [Wed, 24 Mar 2010 17:36:56 +0000 (17:36 +0000)]
MFC r205161:
It seems PCI_OUR_REG_[1-5] registers are not mapped on PCI
configuration space on Yukon Ultra(88E8056) such that accesses to
these registers were NOPs which in turn make msk(4) instable on
this controller. Use indirect access method to access
PCI_OUR_REG_[1-5] registers. This should fix a long standing
instability bug which prevented msk(4) working on Yukon Ultra.
Special thanks to koitsu who gave me remote access to his system.
yongari [Wed, 24 Mar 2010 17:29:32 +0000 (17:29 +0000)]
MFC r204975,204978-204979,204981:
r204975:
Enable hardware fixes for BCM5704 B0 as recommended by data sheet.
r204978:
Set maximum read byte count to 2048 for PCI-X BCM5703/5704 devices.
Also disable relaxed ordering as recommended by data sheet for
PCI-X devices. For PCI-X BCM5704, set maximum outstanding split
transactions to 0 as indicated by data sheet.
For BCM5703 in PCI-X mode, DMA read watermark should be less than
or equal to maximum read byte count configuration. Enforce this
limitation in DMA read watermark configuration.
r204979:
Fix typo in r204975.
r204981:
Fix typo in r204978.
yongari [Wed, 24 Mar 2010 17:18:44 +0000 (17:18 +0000)]
MFC r204545:
Remove taskqueue based interrupt handling. After r204541 msk(4)
does not generate excessive interrupts any more so we don't need
to have two copies of interrupt handler.
While I'm here remove two STAT_PUT_IDX register accesses in LE
status event handler. After r204539 msk(4) always sync status LEs
so there is no need to resort to reading STAT_PUT_IDX register to
know the end of status LE processing. Just trust status LE's
ownership bit.
yongari [Wed, 24 Mar 2010 17:11:01 +0000 (17:11 +0000)]
MFC r204541:
Implement rudimentary interrupt moderation with programmable
countdown timer register. The timer resolution may vary among
controllers but the value would be represented by core clock
cycles. msk(4) will automatically computes number of required clock
cycles from given micro-seconds unit.
The default interrupt holdoff timer value is 100us which will
ensure less than 10k interrupts under load. The timer value can be
changed with dev.mskc.0.int_holdoff sysctl node.
Note, the interrupt moderation is shared resource on dual-port
controllers so you can't use separate interrupt moderation value
for each port. This means we can't stop interrupt moderation in
driver stop routine. Also have msk_tick() reclaim transmitted Tx
buffers as safety belt. With this change there is no need to check
missing Tx completion interrupt in watchdog handler, so remove it.
luigi [Wed, 24 Mar 2010 15:19:47 +0000 (15:19 +0000)]
MFC 205602:
Honor ip.fw.one_pass when a packet comes out of a pipe without being delayed.
I forgot to handle this case when i did the mtag cleanup three months ago.
I am merging immediately because this bugfix is important for
people using RELENG_8.
yongari [Tue, 23 Mar 2010 22:22:26 +0000 (22:22 +0000)]
MFC r204378:
Add TSO support on VLANs. While I'm here remove unnecessary check
of VLAN hardware checksum offloading. vlan(4) already takes care of
this.
yongari [Tue, 23 Mar 2010 22:19:27 +0000 (22:19 +0000)]
MFC r204377:
Add TSO support on VLANs. While I'm here remove unnecessary check
of VLAN hardware checksum offloading. vlan(4) already takes care of
this.
yongari [Tue, 23 Mar 2010 22:16:12 +0000 (22:16 +0000)]
MFC r204376:
Disable TSO on BCM5755M controller until I understand better for
the issue. I still have no idea why TSO does not work on this
controller. davidch@ also confirmed there is no known TSO related
issues for this controller.
yongari [Tue, 23 Mar 2010 22:11:39 +0000 (22:11 +0000)]
MFC r204373-204374:
r204373:
Move TSO setup to new function bce_tso_setup(). Also remove VLAN
parsing code in TSO path as the controller requires VLAN hardware
tagging to make TSO work over VLANs.
While parsing the mbuf in TSO patch, always perform check for
writable mbuf as bce(4) have to reset IP length and IP checksum
field of IP header and make sure to ensure contiguous buffer before
accessing IP/TCP headers. While I'm here replace magic number 40 to
more readable sizeof(struct ip) + sizeof(struct tcphdr).
r204374:
Add TSO support on VLANs. bce(4) controllers require VLAN hardware
tagging to make TSO work on VLANs so explicitly disable TSO on VLAN
if VLAN hardware tagging is disabled.
yongari [Tue, 23 Mar 2010 22:04:18 +0000 (22:04 +0000)]
MFC r204368,204370-204372:
r204368:
Allow disabling VLAN hardware tag stripping with software work
around. Management firmware(ASF/IPMI/UMP) requires the VLAN
hardware tag stripping so don't actually disable VLAN hardware tag
stripping. If VLAN hardware tag stripping was disabled, bce(4)
manually reconstruct VLAN frame by appending stripped VLAN tag.
Also remove unnecessary IFCAP_VLAN_MTU message.
r204370:
Make sure to stop controller first before changing MTU. And if
interface is not running don't initialize controller.
While here remove unnecessary update of error variable.
r204371:
Make toggling TSO, VLAN hardware checksum offloading work. Also fix
TX/RX checksum handler to set/clear relavant assist bits which was
used to cause unexpected results.
With this change, bce(4) can be bridged with other interfaces that
lack TSO, VLAN checksum offloading.
yongari [Tue, 23 Mar 2010 21:51:31 +0000 (21:51 +0000)]
MFC r204363,204365-204367,204539-204540:
r204363:
Optimize inserting LE for TX checksum computation. Controller does
not require checksum LE configuration if checksum start and write
position is the same as before. So keep track last checksum start
and write position and insert new LE whenever the position is
changed. This reduces number of LEs used in TX path as well as
slightly enhance TX performance.
r204365:
Don't hardcod register offset to set PCIe max read request size.
The register offset is not valid on 88E8072 controller. Also don't
blindly increase max read request size to 4096, instead, use 2048
which seems to be more sane value and only change the value if the
hardware default size(512) was used on that register.
For PCIX controllers, use system defined constant rather than using
magic value.
While I'm here stop showing negotiated link width.
r204366:
Allocate single MSI message. msk(4) used to allocate 2 MSI messages
for controllers like 88E8053 which reports two MSI messages.
Because we don't get anything useful things with 2 MSI messages,
allocating 1 MSI message would be more sane approach.
While I'm here, enable MSI for dual-port controllers too. Because
status block is shared for dual-port controllers, I don't think
msk(4) will encounter problem for using MSI on dual-port
controllers.
r204367:
Remove trailing white spaces.
r204539:
Properly sync status LEs after processing.
r204540:
Make sure to enable flow-control only if established link is
full-duplex. Previously msk(4) used to allow flow-control on
1000baseT half-duplex media. Also GMAC pause is enabled if link
partner is capable of handling it.
While I'm here use IFM_OPTIONS instead of using IFM_GMASK to check
optional flags of link.
yongari [Tue, 23 Mar 2010 21:38:25 +0000 (21:38 +0000)]
MFC r204361-204362:
r204361:
Reuse the configured LE for VLAN if new LE was created for TSO.
Only old controllers need to create new LE for TSO. This change
makes TSO work over VLANs.
r204362:
Add TSO support on VLANs. Controller requires VLAN hardware tagging
to make TSO work over VLANs.
yongari [Tue, 23 Mar 2010 19:41:43 +0000 (19:41 +0000)]
MFC r204228,204230:
r204228:
Add TSO support on VLANs. Also make sure to update TSO capability
whenever jumbo frame is configured.
While I'm here remove unnecessary check of VLAN hardware checksum
offloading. vlan(4) already takes care of this.
r204230:
Remove Tx mbuf parsing code for VLAN in TSO path. Controller does
not support TSO over VLAN if VLAN hardware tagging is disabled so
there is no need to check VLAN here.
yongari [Tue, 23 Mar 2010 19:30:15 +0000 (19:30 +0000)]
MFC r204155,204219:
r204155:
Increase PCIe maximuim read request size to 2048. Because re(4) uses
Tx DMA burst size 2048, I beleive PCIe maximum read request size
also should match to the value of Tx DMA burst size. With this
change I can get more than 800Mbps for TCP bulk transfers.
Previously I was not able to get more than 700Mbps. If I enable TSO
it now shows 927Mbps.
r204219:
Add TSO on VLANs. Because re(4) has a TSO limitation for jumbo
frame, make sure to update VLAN capabilities whenever jumbo frame
is configured.
While I'm here rearrange interface capabilities configuration. The
controller requires VLAN hardware tagging to make TSO work on VLANs
so explicitly check this requirement.
yongari [Tue, 23 Mar 2010 19:16:35 +0000 (19:16 +0000)]
MFC r204151,204223:
r204151:
Add TSO support on VLAN. Controller requires VLAN hardware tagging
to make TSO work on VLAN. So if VLAN hardware tagging is disabled
explicitly clear TSO on VLAN. While I'm here remove duplicated
VLAN_CAPABILITIES call.
r204223:
Remove Tx mbuf parsing code for VLAN in TSO path. Controller does
not support TSO over VLAN if VLAN hardware tagging is disabled so
there is no need to check VLAN here.
While I'm here make sure to pullup IP/TCP headers in the first
buffer.
jh [Tue, 23 Mar 2010 16:45:29 +0000 (16:45 +0000)]
MFC r205121:
Use an unique directory name instead of hardcoded /tmp/.diskless.
A malicious user could create a file named /tmp/.diskless and cause
the script to misbehave.
luigi [Tue, 23 Mar 2010 09:58:59 +0000 (09:58 +0000)]
MFC of a large number of ipfw and dummynet fixes and enhancements
done in CURRENT over the last 4 months.
HEAD and RELENG_8 are almost in sync now for ipfw, dummynet
the pfil hooks and related components.
Among the most noticeable changes:
- r200855 more efficient lookup of skipto rules, and remove O(N)
blocks from critical sections in the kernel;
- r204591 large restructuring of the dummynet module, with support
for multiple scheduling algorithms (4 available so far)
See the original commit logs for details.
Changes in the kernel/userland ABI should be harmless because the
kernel is able to understand previous requests from RELENG_8 and
RELENG_7. For this reason, this changeset would be applicable
to RELENG_7 as well, but i am not sure if it is worthwhile.
hrs [Mon, 22 Mar 2010 22:07:19 +0000 (22:07 +0000)]
MFC r203272:
- Fix a bug when adding an interface with an invalid MTU sets the
bridge's MTU if it is the firstly-added one while the addition
itself fails.
- Allow SIOCSIFMTU only when all members have the same MTU.
- Remove IFT_GIF check when defining the brige MTU by the
firstly-added interface's one. The MTU of the gif interface
has to be the same as the bridge's one.
edwin [Mon, 22 Mar 2010 21:35:54 +0000 (21:35 +0000)]
MFC of r205475, tzdata2010f:
The Australian Antartic Division:
- Macquarie Island will stay on UTC+11 for winter and not switch back from DST.
- Casey station reverted to its normal time of UTC+8 on 5 March 2010.
- Davis station will revert to its normal time of UTC+7 at 10 March 2010
- Mawson station stays on UTC+5.
Syria will start DST on Thursday 1 April 2010 at midnight.
jkim [Mon, 22 Mar 2010 20:36:35 +0000 (20:36 +0000)]
MFC: r205223
Fix a long standing regression of readdir(3) in fdescfs(5) introduced
in r1.48. We were stopping at the first null pointer when multiple file
descriptors were opened and one in the middle was closed. This restores
traditional behaviour of fdescfs.
jkim [Mon, 22 Mar 2010 19:59:00 +0000 (19:59 +0000)]
MFC: r205092
Tidy up callout for select(2) and read timeout.
- Add a missing callout_drain(9) before the descriptor deallocation.[1]
- Prefer callout_init_mtx(9) over callout_init(9) and let the callout
subsystem handle the mutex for callout function.
PR: kern/144453
Submitted by: Alexander Sack (asack at niksun dot com)[1]
jhb [Mon, 22 Mar 2010 15:52:26 +0000 (15:52 +0000)]
MFC 204950,205020:
Use thr_once() with once_t controls to initialize various thread_key_t
objects used to provide per-thread storage in the RPC code. Almost all
of these used double-checking with a dedicated mutex (tsd_lock) to do this
before. However, that is not always safe with more relaxed memory orders.
There were also other bugs, such as one in __rpc_createrr() that caused a
new key to be allocated each time __rpc_createrr() was invoked.
nwhitehorn [Sat, 20 Mar 2010 15:27:01 +0000 (15:27 +0000)]
MFC r204297:
Move the OEA64 scratchpage to the end of KVA from the beginning, and set
its PVO to map physical address 0 instead of kernelstart. This fixes a
situation in which a user process could attempt to return this address
via KVM, have it fault while being modified, and then panic the kernel
because (a) it is supposed to map a valid address and (b) it lies in the
no-fault region between VM_MIN_KERNEL_ADDRESS and virtual_avail.
While here, move msgbuf and dpcpu back into regular KVA space for
consistency with other implementations.
nwhitehorn [Sat, 20 Mar 2010 15:23:06 +0000 (15:23 +0000)]
MFC r204296:
Provide an implementation of pmap_dev_direct_mapped() on OEA64. This is
required in order to be able to mmap the running kernel, which is turn
required to avoid fstat returning gibberish.
nwhitehorn [Sat, 20 Mar 2010 15:19:57 +0000 (15:19 +0000)]
MFC r204268:
Close a race involving the OEA64 scratchpage. When the scratch page's
physical address is changed, there is a brief window during which its PTE
is invalid. Since moea64_set_scratchpage_pa() does not and cannot hold
the page table lock, it was possible for another CPU to insert a new PTE
into the scratch page's PTEG slot during this interval, corrupting both
mappings.
Solve this by creating a new flag, LPTE_LOCKED, such that
moea64_pte_insert will avoid claiming locked PTEG slots even if they
are invalid. This change also incorporates some additional paranoia
added to solve things I thought might be this bug.
nwhitehorn [Sat, 20 Mar 2010 15:15:54 +0000 (15:15 +0000)]
MFC r204128:
Reduce KVA pressure on OEA64 systems running in bridge mode by mapping
UMA segments at their physical addresses instead of into KVA. This emulates
the direct mapping behavior of OEA32 in an ad-hoc way. To make this work
properly required sharing the entire kernel PMAP with Open Firmware, so
ofw_pmap is transformed into a stub on 64-bit CPUs.
Also implement some more tweaks to get more mileage out of our limited
amount of KVA, principally by extending KVA into segment 16 until the
beginning of the first OFW mapping.
nwhitehorn [Sat, 20 Mar 2010 15:05:44 +0000 (15:05 +0000)]
Fix a bug where pages being removed from memory entirely no longer have
PVOs, and so the modified state of the page can no longer be communicated
to the VM layer, causing pages not to be flushed to swap when needed, in
turn causing memory corruption. Also make several correctness adjustments
to I-Cache synchronization and TLB invalidation for 64-bit Book-S CPUs.
nwhitehorn [Sat, 20 Mar 2010 14:55:22 +0000 (14:55 +0000)]
MFC r204903:
Place interrupt handling in a critical section and remove double
counting in incrementing the interrupt nesting level. This fixes a number
of bugs in which the interrupt thread could be preempted by an IPI,
indefinitely delaying acknowledgement of the interrupt to the PIC, causing
interrupt starvation and hangs.
Provide thermal management and monitoring features in smu(4). This allows
fan control and thermal monitoring on SMU-based Apple G5 machines, as well
as an led(4) interface to control the sleep LED.
yongari [Fri, 19 Mar 2010 00:26:45 +0000 (00:26 +0000)]
MFC r202717:
- Added a workaround for NC-SI management firmware that would allow
frames to be accepted while the driver is resetting the hardware.
This failure is generally observed when broadcast frames are received
during driver load and will generate "Unable to write CTX memory"
errors.
- Small changes to driver flags display.
yongari [Thu, 18 Mar 2010 19:04:04 +0000 (19:04 +0000)]
MFC r204149:
Add TSO support on VLANs. Intentionally separated IFCAP_VLAN_HWTSO
from IFCAP_VLAN_HWTAGGING. I think some hardwares may be able to
TSO over VLAN without VLAN hardware tagging.
Driver changes and userland support will follow.
yongari [Thu, 18 Mar 2010 18:50:20 +0000 (18:50 +0000)]
MFC r203358,203716:
r203358:
PCI express device status register has W1C feature. Writing 0 has
no effect. Make sure to clear error bits by writing 1. [1]
While I'm here use predefined value instead of hardcodig magic
vlaue.
Submitted by: msaitoh at NetBSD [1]
r203716:
Move device specific flag configuration to attach routine.
The softc obtained in device probe wouldn't be the same one used in
device attach. Drivers should not assume any values stored in softc
structure in probe routine will be available for its attach routine.
yongari [Thu, 18 Mar 2010 18:35:28 +0000 (18:35 +0000)]
MFC r202821-202822.
r202821:
Fix a long standing ASF heartbeat sending bug. The initial
implementation of heartbeat interval was 2 but there was typo which
caused the heartbeat is sent approximately every 5 seconds. This
caused unintended controller reset by firmware because firmware
thought OS was crashed.
Submitted by: Floris Bos < info <> je-eigen-domein dot nl >
Tested by: Andrzej Tobola < ato <> iem dot pw dot edu dot pl >
r202822:
Use new handshake command for BCM5750 or new controllers.
gavin [Wed, 17 Mar 2010 20:39:21 +0000 (20:39 +0000)]
Merge r203622 from head:
Add support for a few more Sony-specific ACPI features (default display
brightness, wired LAN power and bass gain), and update the description of
one previously unknown feature (display contrast). While here, expand on
a comment and remove two defines left over from an old version of the code.
Also update man page to document the above changes, and correct grammar.
gavin [Wed, 17 Mar 2010 20:27:35 +0000 (20:27 +0000)]
Merge r203835 from head:
When growing a UFS1 filesystem, we need to initialise all inodes in any new
cylinder groups that are created. When the filesystem is first created,
newfs always initialises the first two blocks of inodes, and then in the
UFS1 case will also initialise the remaining inode blocks. The changes in
growfs.c 1.23 broke the initialisation of all inodes, seemingly based on
this implementation detail in newfs(8). The result was that instead of
initialising all inodes, we would actually end up initialising all but the
first two blocks of inodes. If the filesystem was grown into empty
(all-zeros) space then the resulting filesystem was fine, however when
grown onto non-zeroed space the filesystem produced would appear to have
massive corruption on the first fsck after growing.
A test case for this problem can be found in the PR audit trail.
Fix this by once again initialising all inodes in the UFS1 case.
dougb [Wed, 17 Mar 2010 07:26:00 +0000 (07:26 +0000)]
MFC r205145:
Make it more clear in the docs that -a is not compatible with -iFU,
and enforce this in the code. Apparently a lot of users mistakenly
combine -a with these flags and are then mystified that no changes
were made.
While I'm here, fix a trailing space in mergemaster.8
marius [Mon, 15 Mar 2010 19:13:36 +0000 (19:13 +0000)]
MFC: r204222
According to the Linux sungem driver, in case of Apple (K2) GMACs
GEM_MIF_CONFIG_MDI0 cannot be trusted when the firmware has powered
down the chip so the internal transceiver has to be hardcoded. This
is also in line with the AppleGMACEthernet driver, which just doesn't
distinguish between internal/external transceiver and MDIO/MDI1
respectively in the first place. Tested by: Andreas Tobler
marius [Mon, 15 Mar 2010 18:32:57 +0000 (18:32 +0000)]
MFC: r204974, r205002
- The OPSZ macro actually only does the right thing for int32 and int64
operands but not for double and extended double ones. Instead of trying
to fix the macro just nuke it and unroll the loops in the correct way
though as extended double operands turn out to be the only special case.
- For FxTO{s,d,q} the source operand is int64 so rs2 has to be re-decoded
after setting type accordingly as it's generally decoded using the low
2 bits as the type, which are 0 for these three instructions.
- Similarly, in case of F{s,d,q}TOx the target is int64 so rd has to be
re-decoded using not only the operand mask appropriate for int64 but
also the correct register number encoding.
- Use const where appropriate.
- Wrap long lines.
jilles [Sun, 14 Mar 2010 13:07:40 +0000 (13:07 +0000)]
MFC r204410: Include terminated threads in ps's process cpu time field.
When a kinfo_proc is filled, first fill_kinfo_proc_only() fills in
ki_runtime using p->p_rux.rux_runtime (all cpu time used by the process
including terminated threads). If information for a specific thread is
requested, fill_kinfo_thread() then overwrites this with the thread's
td->td_runtime (good). If not, fill_kinfo_aggregate() overwrote it with
the sum of all threads' td->td_runtime which does not include terminated
threads.
This affects ps(1)'s TIME field, not its %CPU field nor anything in
top(1).
jh [Fri, 12 Mar 2010 06:56:51 +0000 (06:56 +0000)]
MFC r204447:
In _gettemp(), check that the length of the path doesn't exceed
MAXPATHLEN. Otherwise the path name (or part of it) may not fit to
carrybuf causing a buffer overflow.
delphij [Fri, 12 Mar 2010 00:51:13 +0000 (00:51 +0000)]
MFC r203760: Improve time precision for grdc(6):
Traditionally, grdc would obtain time through time(3) which in turn gets
only the second part of clock (CLOCK_SECOND), and sleep for 1 second after
each screen refresh.
This approach would have two problems. First, we are not guaranteed to
be waken up at the beginning of a whole second, which will typically
exhibit as a "lag" on second number. Second, because we sleep for whole
second, and the refresh process would take some time, the error would
accumulate from time to time, making the lag variable.
Make grdc(6) to use time(3) to get time only at the beginning, and sample
time in CLOCK_REALTIME_FAST granularity after refreshing, and use the
nanosecond part to caculate how much time we want to sleep.
projects / FreeBSD / stable / 8.git / log