MFV r297760: 6418 zpool should have a label clearing command
Reviewed by: Yuri Pankov <yuri.pankov@nexenta.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: George Wilson <george.wilson@delphix.com>
Author: Will Andrews <will@firepipe.net>
ian [Sat, 9 Apr 2016 19:09:06 +0000 (19:09 +0000)]
Align the start of the text segment to an 8-byte boundary. This fixes
alignment aborts in ubldr.bin for RPi that started happening with clang 3.8
(earlier clang apparently didn't generate strd instructions that trigger
the alignment fault). The abort happened in ubldr.bin and not ubldr (elf
version) because the elf headers are 0xf4 bytes long, and stripping them
off left everything 4-byte aligned.
While here, also stop aligning the data segment to a page boundary, align
it to 8 bytes instead (aligning to a page just needlessly makes the file
bigger); pointed out by andrew@.
Add more fine-grained kernel options for NUMA support.
VM_NUMA_ALLOC is used to enable use of domain-aware memory allocation in
the virtual memory system. DEVICE_NUMA is used to enable affinity
reporting for devices such as bus_get_domain().
MAXMEMDOM must still be set to a value greater than for any NUMA support
to be effective. Note that 'cpuset -gd' always works if MAXMEMDOM is
enabled and the system supports NUMA.
It looks like as with the safety belt of DELAY() fastened (*) we can
completely tear down and free all memory for TCP (after r281599).
(*) in theory a few ticks should be good enough to make sure the timers
are all really gone. Could we use a better matric here and check a
tcbcb count as an optimization?
PR: 164763
Reviewed by: gnn, emaste
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D5734
We attach the "counter" to the tcpcbs. Thus don't free the
TCP Fastopen zone before the tcpcbs are gone, as otherwise
the zone won't be empty.
With that it should be safe to destroy the "tfo" zone without
leaking the memory.
PR: 164763
Reviewed by: gnn
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D5731
While there is no dependency interaction, stopping the timer before
freeing the rest of the resources seems more natural and avoids it
being scheduled an extra time when it is no longer needed.
Reviewed by: gnn, emaste
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D5733
Make the KASSERT message in hash destroy more informative.
While the pointer might not be too helpful, the malloc type might at
least give a good hint about which hashtbl we are talking.
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Reviewed by: gnn, emaste
Differential Revision: https://reviews.freebsd.org/D5802
adrian [Sat, 9 Apr 2016 00:58:38 +0000 (00:58 +0000)]
[ath] Only process beacon frames for the IBSS/BSSID if appropriate.
* Don't use arbitrary frames for the average RX RSSI - only frames
from the current BSSID
* Don't log / do the syncbeacon logic for another BSSID and definitely
don't do the syncbeacon call if we miss beacons outside of STA mode.
* Don't do the IBSS merge bits if the current node plainly won't ever
match our current BSS (ie, the IBSS doesn't have to match, but all
the same bits that we check in ieee80211_ibss_merge() have to match.)
Tested:
* ath(4), AR9380, IBSS mode, surrounded by a lot of IBSS 11ac networks.
adrian [Sat, 9 Apr 2016 00:55:55 +0000 (00:55 +0000)]
[net8021] Pull out the ibss check code into a public function.
The ath(4) driver now sees beacons and management frames for different
BSSIDs in IBSS mode, which is a problem when you're in a very busy
IBSS environment.
So, expose this function so drivers can use it to check if the current
RX node is actually for a BSS we need to pay attention to or not.
PR: kern/208644
Sponsored by: Eva Automation. Inc.
This revision introduces PCIe support for the relevant Mediatek/Ralink
SoCs.
Currently the PCIe support is not converted to INTRNG, this may be a
task for the future.
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5886
Disable USB PHY slew rate calibration for Mediatek SoCs for now
USB on both MT7621 and MT7688 seems to work much better without doing
slew rate calibration.
These are the only two SoCs, apart from MT7628, which actually make
use of the slew rate calibration routines implemented in the mtk_usb_phy
driver. Since MT7628 is actually a superset of MT7688 things should be
the same for it as well.
We do not remove the code, we simply define it out.
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5884
zio write issue threads should have lower (numerically greater) priority
This is because they might do data compression which is quite CPU
expensive. The original code is correct for illumos, because there
a higher priority corresponds to a greater number.
This is rather ugly, because the RIPE and APNIC whois servers do
not provide referrals for address blocks that they do not manage.
However ARIN is usually the right place or knows the right place
so we try there.
The particular instance which clued me in to this bug is U.Mich.
141.211.0.0/16 for which the referral chain should be IANA ->
RIPE -> ARIN. RIPE's RDAP does provide useful redirects (for
example try `curl -I http://rdap.db.ripe.net/ip/141.211.0.0)
so maybe their whois server can be improved.
AfriNIC's whois server gives more direct referrals, but they are
designed to be human-readable. Ugly, but we can manage.
The issue of referrals between RIRs is likely to become more important
in the future whith the increasing number of cross-region IP address
block transfers increases.
The previous values caused the callout thread stall for 100ms each 2s
if no link is present. Dtrace analysis showed that it has significant
impact on overall interrupt performance.
Decrease these values by a factor of 100.
Currently we don't keep zoneid in in6_ifaddr structure, because there
is still some code, that doesn't properly initialize sin6_scope_id,
but some functions use sa_equal() for addresses comparison. sa_equal()
compares full sockaddr_in6 structures and such comparison will fail.
For now use zero zoneid in in6ifa_ifwithaddr(). It is safe, because
used address is in embedded form. In future we will use zoneid, so mark it
with XXX comment.
Add function for mapping SRAM-D area to USB0 (OTG) controller. Use a lower
pass number to ensure that this driver is loaded before EMAC or OTG,
regardless of the order of nodes in the DT.
aq64_minfree is unsigned so comparing to find out if it is less
than zero is a nonsense. Move the comparison to the last position
as we don't want to spend time if any of the others triggers first.
hile it would be tempting to just remove it, it may be important to
keep it for portability with platforms where may be signed(?) or
in case we may want to change it in the future.
This flag indicates that the user wishes to use the GELIBOOT feature to boot from a fully encrypted root file system.
Currently, GELIBOOT does not support key files, and in the future when it does, they will be loaded differently.
Due to the design of GELI, and the desire for secrecy, the GELI metadata does not know if key files are used or not, it just adds the key material (if any) to the HMAC before the optional passphrase, so there is no way to tell if a GELI partition requires key files or not.
Since the GELIBOOT code in boot2 and the loader does not support keys, they will now only attempt to attach if this flag is set. This will stop GELIBOOT from prompting for passwords to GELIs that it cannot decrypt, disrupting the boot process
Improve over the solution in r297527:
Instead of attempting to initialize all the possible cases, just
move the check nearer to the case where it makes sense.
ian [Thu, 7 Apr 2016 19:51:27 +0000 (19:51 +0000)]
Remove unecessary locking, mostly from places where a read is done of a
value that can't ever be in an inconsistant intermediate state even when
some other thread is in the middle of writing the value/register.
Locking of the hardware remains in the few places that do r-m-w operations.
Locking of metadata access is restricted to places using memcpy or sprintf
to modify the metadata.
ian [Thu, 7 Apr 2016 19:17:47 +0000 (19:17 +0000)]
Code cleanup: stop searching for a pin in the array and just use the pin
number directly as an index. We create the array ourselves and nothing
can change the order of items in it, it's a simple 1:1 mapping.
ian [Thu, 7 Apr 2016 18:19:09 +0000 (18:19 +0000)]
Fix a copyright glitch before it gets copy-pasted again. I think this must
have started as collateral damage in a global search-replace, then it got
copied around when I cloned a file to begin creating a new file.
ian [Thu, 7 Apr 2016 17:45:01 +0000 (17:45 +0000)]
Comestic changes; when INTRNG support was added, some functions became
oddly separated from related functionality. This just moves some blocks
of code around so that setup_intr and teardown_intr are near each other
again, and likewise for enable/disable_intr. No functional changes.
Associate device_t objects with ACPI handles via PCI_CHILD_ADDED().
Previously, the ACPI PCI bus driver did a single pass over the devices in
the namespace that were a child of a given PCI bus to associate the
PCI bus-enumerated device_t devices with the corresponding ACPI handles.
However, this meant that handles were only established at runtime for devices
found during the initial PCI bus scan.
PCI_IOV adds devices that show up after the initial PCI bus scan, and coming
changes to add a bus rescan can also add devices after the initial scan.
This change adds a pci_child_added() callback to the ACPI PCI bus that walks
the namespace to find the ACPI handle for each device that is added. Using
a callback means that the handle is correctly set for any device no matter
how it is added (initial scan, IOV, or a bus rescan).
Fix intr_irq_shuffle(). After r297539, ISRCs doing IPI may be also
registered into global interrupt table. Thus, they must be filtered out
like per-cpu interrupts. Fortunately, it does not influence anything
on interrupt controllers which already use INTRNG.
Initial import of Ralink/Mediatek MIPS SoC support #7
GPIO controller drivers import.
As with other Ralink/Mediatek work, there are 2 versions of the GPIO
controller driver, depending on the type of SoC.
This revision introduces initial support for these.
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5877
Implement intr_isrc_init_on_cpu() and use it to replace very same
code implemented in every interrupt controller driver running SMP.
This function returns true, if provided ISRC should be enabled on
given cpu.
Alike to r293708 relax pool check in vdev_geom_open_by_path().
This made impossible spare disk open by known path, which kind of worked
only because the same fix was applied to vdev_geom_attach_by_guids() in
r293708.
Initial import of Ralink/Mediatek MIPS SoC support #6
SPI drivers for the various Ralink/Mediatek SoCs. There are 2 versions of
the SPI controller (so far) present in the supported SoCs, hence v1 and v2
drivers.
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5842
Initial import of Ralink/Mediatek MIPS SoC support #5
USB support
This revision adds USB (EHCI/OHCI/OTG, depending on SoC type) support for
various Ralink/Mediatek SoCs.
Currently USB is not supported on MT7621, this will be a future addition.
A USB PHY driver is also included, so that we can properly initialize the
USB PHY (e.g., clocks, resets, registers where needed), before attempting
to initialize EHCI/OHCI/OTG functionality.
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5841
Initial import of Ralink/Mediatek MIPS SoC support #4
UART drivers.
- uart_dev_mtk.[ch] are the old-style Mediatek/Ralink-specific UART driver
as also found in sys/mips/rt305x/uart_dev_rt305x.c, with minor improvements
and FDT attachment enabled for the appropriate SoCs.
- uart_dev_mtk_ns8250.c is the new-style ns16550a-compatible UART driver
found in newer Mediatek SoCs. It uses the uart_dev_ns8250.c driver
indirectly and is basically just a wrapper around it and only overrides its
probe method.
The reason I am not using the uart_dev_ns8250.c driver directly is because
I have some code that does UART clock detection before initializing the
UART, so that we don't need to hard-code the UART clock frequency in the
dts files for each board.
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5840
Initial import of Ralink/Mediatek MIPS SoC support #3
Interrupt controllers found in various Mediatek/Ralink SoCs.
mtk_intr_v1 and mtk_intr_v2 are basically the same at the moment, with
just different register mappings.
However, v1 interrupt controller has a subset of the functionality of the
v2 interrupt controller, so in the future the v2 interrupt controller driver
may be enhanced, if needed, with things like level/edge interrupts and soft
interrupts. So, for the moment I suggest we keep them as 2 separate files.
mtk_intr_gic provides very basic (similar to v1 and v2) support for MIPS GIC
controllers, which currently maps all interrupts to a single core and sets
them to type level, active high. In the future this may be developed into a
generic GIC controller to support any new MIPS SoCs that include it. The GIC
is a standard MTI interrupt controller in their multi-core line-up (e.g.,
1004K, 1074K, etc.), rather than a SoC-specific controller.
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5839
Initial import of Ralink/Mediatek MIPS SoC support #2
This revision adds the following to the Mediatek/Ralink support:
- initial support for "clocks" FDT property, currently based on fdt_clock
- initial support for "resets" FDT property, currently based on the
fdt_reset interface from D5826
- initial support for "pinctrl,bits" functionality via FDT. May be extended
in the future to cover a better and fuller pinctrl implementation
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5827
- machine dependent low level init code
- SoC clocks detection and some utility functions
- Common interface to read/write/modify SoC system control registers, used
by some of the other drivers and utility functions
- simple FDT resets support, based on the fdt_clock implementation already
in the tree. For the moment resets and clocks are managed using these
implementations. I am planning to port those to the new extres framework
in the future, but currently I simply don't have time to do this part too.
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5826
Fix interrupts delivery on ThunderX for VF IDs beyond 8
SR-IOV devices usually use Alternative Routing ID (ARI).
In that case slot/device is always assumed to be 0 and
function/identifier is extended to 8 bits.
Fix interrupts delivery to VF IDs beyond 8 by using a correct
DevID if ARI is enabled.
A couple of minor changes that I missed that Michael had done, most noted
in these is the change to non-strict ordering for incoming data (this will
make pkt-drill test 14 fail but its expected).
This is work done by Michael Tuexen and myself at the IETF. This
adds the new I-Data (Interleaved Data) message. This allows a user
to be able to have complete freedom from Head Of Line blocking that
was previously there due to the in-ability to send multiple large
messages without the TSN's being in sequence. The code as been
tested with Michaels various packet drill scripts as well as
inter-networking between the IETF's location in Argentina and Germany.
This revision adds support to if_rt for more SoCs.
The SoCs I've tried the driver with include the following:
RT3050, RT5350, RT3662, RT3883, MT7620, MT7621, MT7688.
On boards, based on the above SoCs traffic is passing through correctly
and the boards survive a flood ping with very little or no drops (drops
may be caused elsewhere in my test setup, however).
One issue still remains and needs to be fixed in the future: if_rt does
not survive an ifconfig rt0 down/ifconfig rt0 up cycle.
This issue existed before this commit as well, however.
Reviewed by: ray
Approved by: adrian (mentor)
Sponsored by: Smartcom - Bulgaria AD
Differential Revision: https://reviews.freebsd.org/D5864
hyperv: Use lapic_{alloc,free}_ipi to allocate private interrupt vector
Suggested by: jhb
Reviewed by: Dexuan Cui <decui microsoft com>, Jun Su <junsu microsoft com>
Sponsored by: Microsoft OSTC
Differential Revision: https://reviews.freebsd.org/D5850
Disable the NetBSD-specific EFAULT requirements test in gettimeofday_err
FreeBSD doesn't specifically list this as a supported error, and in some
configurations/versions of FreeBSD, this test will segfault as the memory
address might be evaluated in userspace, instead of in kernel space like
in NetBSD.
hyperv/vmbus: Use default mtx for channel message queue
First of all sema_post() can't be called w/ spinlock, and the channel
message queue processing is not on hot code path, i.e. spinlock is not
necessary.
Submitted by: Jun Su <junsu microsoft com>
Reviewed by: sephe, Dexuan Cui <decui microsoft com>
MFC after: 1 week
Sponsored by: Microsoft OSTC
Differential Revision: https://reviews.freebsd.org/D5812