With the arrival of 128+ cores it is necessary to handle more than that.
One of the first thing to change is the support for cpumask_t that needs
to handle more than 32 bits masking (which happens now). Some places,
however, still assume that cpumask_t is a 32 bits mask.
Fix that situation by using always correctly cpumask_t when needed.
While here, remove the part under STOP_NMI for the Xen support as it
is broken in any case.
Additively make ipi_nmi_pending as static.
Reviewed by: jhb, kmacy
Tested by: Giovanni Trematerra <giovanni dot trematerra at gmail dot com>
- For CPUs that only support MCE (the machine check exception) but not MCA
(i.e. Pentium), all this does is print out the value of the machine check
registers and then panic when a machine check exception occurs.
- For CPUs that support MCA (the machine check architecture), the support is
a bit more involved.
- First, there is limited support for decoding the CPU-independent MCA
error codes in the kernel, and the kernel uses this to output a short
description of any machine check events that occur.
- When a machine check exception occurs, all of the MCx banks on the
current CPU are scanned and any events are reported to the console
before panic'ing.
- To catch events for correctable errors, a periodic timer kicks off a
task which scans the MCx banks on all CPUs. The frequency of these
checks is controlled via the "hw.mca.interval" sysctl.
- Userland can request an immediate scan of the MCx banks by writing
a non-zero value to "hw.mca.force_scan".
- If any correctable events are encountered, the appropriate details
are stored in a 'struct mca_record' (defined in <machine/mca.h>).
The "hw.mca.count" is a count of such records and each record may
be queried via the "hw.mca.records" tree by specifying the record
index (0 .. count - 1) as the next name in the MIB similar to using
PIDs with the kern.proc.* sysctls. The idea is to export machine
check events to userland for more detailed processing.
- The periodic timer and hw.mca sysctls are only present if the CPU
supports MCA.
Discussed with: emaste (briefly)
MFC after: 1 month
introduced in amd64 revision 1.540 and i386 revision 1.547. However, it
had no harmful effects until after a recent change, r189698, on amd64.
(In other words, the error is harmless in RELENG_7.)
The error is triggered by the failure to allocate a pv entry for the one
and only mapping in a page table page. I am addressing the error by
changing pmap_copy() to abort if either pv entry allocation or page
table page allocation fails. This is appropriate because the creation of
mappings by pmap_copy() is optional. They are a (possible) optimization,
and not a requirement.
Correct a nearby whitespace error in the i386 pmap_copy().
Crash reported by: jeff@
MFC after: 6 weeks
to 2.4.0, as it has appeared in the 2.4.0-rc7 first time.
Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK),
glibc falls back to the hard-coded CLK_TCK value when aux entry
is not present.
Glibc versions prior to 2.2.1 always use hard-coded CLK_TCK value.
For older applications/libc's which depends on hard-coded CLK_TCK
value user should set compat.linux.osrelease less than 2.4.0.
Approved by: kib (mentor)
The frequency of the statistics clock is given by stathz.
Use stathz if it is available, otherwise use hz.
Pointed out by: bde
Approved by: kib (mentor)
I really don't want any pieces of code to include ioctl_compat.h, so let
the ibcs2 and svr4 compat leave sgtty alone. If they want to support
sgtty, they should emulate it on top of termios, not sgtty.
The code has been marked with BURN_BRIDGES for a long time. ibcs2 and
svr4 are not really popular pieces of code anyway.
virtualized instances of hostname and domainname, as well as a new top-level
virtualization struct vimage, which holds pointers to struct vnet and struct
vprocg. Struct vprocg is likely to become replaced in the near future with
a new jail management API import.
As a consequence of this change, change struct ucred to point to a struct
vimage, instead of directly pointing to a vnet.
Merge vnet / vimage / ucred refcounting infrastructure from p4 / vimage
branch.
Permit kldload / kldunload operations to be executed only from the default
vimage context.
This change should have no functional impact on nooptions VIMAGE kernel
builds.
Reviewed by: bz
Approved by: julian (mentor)
OSD-based jail extensions. This allows the Linux MIB to accessed via
jail_set and jail_get, and serves as a demonstration of adding jail support
to a module.
Reviewed by: dchagin, kib
Approved by: bz (mentor)
fix SMP topology detection. On i386, we extend it to cover Core, Core 2,
and Core i7 processors, not just Pentium 4 family, and move it to better
place. On amd64, all supported Intel CPUs should have this MSR.
and hide it inside of atrtc driver. Add new tunable hint.atrtc.0.clock
controlling it. Setting it to 0 disables using RTC clock as stat-/
profclock sources.
Teach i386 and amd64 SMP platforms to emulate stat-/profclocks using i8254
hardclock, when LAPIC and RTC clocks are disabled.
This allows to reduce global interrupt rate of idle system down to about
100 interrupts per core, permitting C3 and deeper C-states provide maximum
CPU power efficiency.
Restore previous behaviour for the case of unknown interrupt. Invocation
of IRQ -1 crashes my system on resume. Returning 0, as it was, is not
perfect also, but at least not so dangerous.
IRQ0 routing on LAPIC-enabled systems.
Add hint.apic.0.clock tunable. Setting it 0 disables using LAPIC timers
as hard-/stat-/profclock sources falling back to using i8254 and rtc timers.
On modern CPUs LAPIC is a part of CPU core which is shutting down when CPU
enters C3 or deeper power state. It makes no problems for interrupt
processing, as chipset wakes up CPU on interrupt triggering. But entering
C3 state kills LAPIC timer and freezes system time, making C3 and deeper
states practically unusable. Using i8254 timer allows to avoid this
problem.
By using i8254 timer my T7700 C2D CPU with UP kernel successfully enters
C3 state, saving more then a Watt of total idle power (>10%) in addition to
all other power-saving techniques.
This technique is not working for SMP yet, as only one CPU receives
timer interrupts. But I think that problem could be fixed by forwarding
interrupts to other CPUs with IPI.
Old implemention used Giant to protect the kernel data structures,
but at the same time called malloc(M_WAITOK), that could cause the
calling thread to sleep and lost Giant protection. User-visible
result was the missed wakeup.
New implementation uses one sx lock per futex. The sx protects
the futex structures and allows to sleep while copyin or copyout
are performed.
Unlike linux, we return EINVAL when FUTEX_CMP_REQUEUE operation
is requested and either caller specified futexes are equial or
second futex already exists. This is acceptable since the situation
can only occur from the application error, and glibc falls back to
old FUTEX_WAKE operation when FUTEX_CMP_REQUEUE returns an error.
Approved by: kib (mentor)
MFC after: 1 month
- Avoid possible divide-by-zero panic on SMP system when the CPUID is
disabled, unsupported, or buggy.
Submitted by: pluknet (pluknet at gmail dot com)[1]
topology of nehalem/corei7 based systems.
- Remove the cpu_cores/cpu_logical detection from identcpu.
- Describe the layout of the system in cpu_mp_announce().
Sponsored by: Nokia
used in some cases):
- Ignore DMA tag boundaries when allocating bounce pages. The boundaries
don't determine whether or not parts of a DMA request bounce. Instead,
they are just used to carve up segments.
- Allow tags with sub-page alignment to share bounce pages since bounce
pages are always page aligned.
Reviewed by: scottl (amd64)
MFC after: 1 month
logical CPUs in a package. We do this by numbering the non-boot CPUs
by starting with the first CPU whose APIC ID is after the boot CPU and
wrapping back around to APIC ID 0 if needed rather than always starting
at APIC ID 0. While here, adjust the cpu_mp_announce() routine to list
CPUs based on the mapping established by assign_cpu_ids() rather than
making assumptions about the algorithm assign_cpu_ids() uses.
MFC after: 1 month
a fair number of static data structures, making this an unlikely
option to try to change without also changing source code. [1]
Change default cache line size on ia64, sparc64, and sun4v to 128
bytes, as this was what rtld-elf was already using on those
platforms. [2]
Suggested by: bde [1], jhb [2]
MFC after: 2 weeks
CACHE_LINE_SIZE constant. These constants are intended to
over-estimate the cache line size, and be used at compile-time
when a run-time tuning alternative isn't appropriate or
available.
Defaults for all architectures are 64 bytes, except powerpc
where it is 128 bytes (used on G5 systems).
MFC after: 2 weeks
Discussed on: arch@
as well as providing stateful load balancing when used with RADIX_MPATH.
- Currently compiled in to i386 and amd64 but disabled by default, it can be enabled at
runtime with 'sysctl net.inet.flowtable.enable=1'.
- Embedded users can remove it entirely from the kernel by adding 'nooption FLOWTABLE' to
their kernel config files.
- A minimal hookup will be added to ip_output in a subsequent commit. I would like to see
more review before bringing in changes that require more churn.
Supported by: Bitgravity Inc.
a tag that has BUS_DMA_KEEP_PG_OFFSET set. Otherwise the page could be
reused with a non-zero offset by a tag that doesn't have
BUS_DMA_KEEP_PG_OFFSET leading to data corruption.
Sleuthing by: avg
Reviewed by: scottl
naming of the partitions (GEOM_PART_EBR_COMPAT). When
compatibility is enabled, changes to the partitioning are
disallowed.
Remove the device name aliasing added previously to provide
backward compatibility, but which in practice doesn't give
us anything.
Enable compatibility on amd64 and i386.
- Do not iterate int 15h, function e820h twice. Instead, we use STAILQ to
store each return buffer and copy all at once.
- Export optional extended attributes defined in ACPI 3.0 as separate
metadata. Currently, there are only two bits defined in the specification.
For example, if the descriptor has extended attributes and it is not
enabled, it has to be ignored by OS. We may implement it in the kernel
later if it is necessary and proven correct in reality.
- Check return buffer size strictly as suggested in ACPI 3.0.
Reviewed by: jhb
Remove a hack to generate more efficient code for port numbers below
0x100, which has been obsolete for at least ten years, because GCC has
an asm constraint to specify that.
Submitted by: Christoph Mallon <christoph mallon gmx de>
Most compilers nowadays (including GCC) are smart enough to know what's
going on and generate more efficient code anyway.
Submitted by: Christoph Mallon <christoph.mallon@gmx.de>
1) Move the new field (brand_note) to the end of the Brandinfo structure.
2) Add a new flag BI_BRAND_NOTE that indicates that the brand_note pointer
is valid.
3) Use the brand_note field if the flag BI_BRAND_NOTE is set and as old
modules won't have the flag set, so the new field brand_note would be
ignored.
Suggested by: jhb
Reviewed by: jhb
Approved by: kib (mentor)
MFC after: 6 days
but I see no benefit from it today.
VM_PROT_READ_IS_EXEC was only intended for use on processors that do not
distinguish between read and execute permission. On an mmap(2) or
mprotect(2), it automatically added execute permission if the caller
specified permissions included read permission. The hope was that this
would reduce the number of vm map entries needed to implement an address
space because there would be fewer neighboring vm map entries that differed
only in the presence or absence of VM_PROT_EXECUTE. (See vm/vm_mmap.c
revision 1.56.)
Today, I don't see any real applications that benefit from
VM_PROT_READ_IS_EXEC. In any case, vm map entries are now organized
as a self-adjusting binary search tree instead of an ordered list. So,
the need for coalescing vm map entries is not as great as it once was.
To keep these structures ABI-compatible, half the size of r_trapno,
r_err, mc_trapno, mc_flags.
Add fsbase and gsbase to mcontext on both amd64 and i386.
Add flags to amd64 mcontext to indicate that it contains valid segments
or bases.
In collaboration with: pho
Discussed with: peter
Reviewed by: jhb
as 'real memory' instead of Maxmem if the value is available.
Note amd64 displayed physmem as 'usable memory' since machdep.c r1.640
to unconfuse users. Now it is consistent across amd64 and i386 again.
While I am here, clean up smbios.c a bit and update copyright date.
Reviewed by: jhb
in AMD FPUs:
- Do not clear the affected state in the case that the FPU registers for
the thread that already owns the FPU are changed via fpu_setregs(). The
only local information the thread would see is its own state in that
case.
- Fix a type mismatch for the dummy variable used in a "fld". It accepts
a float, not a double.
Reviewed by: bde
Approved by: so (cperciva)
MFC after: 1 month
the requested PCI bus falls outside of the bus range given in the ACPI
MCFG table. Several BIOSes seem to not include all of the PCI busses in
systems in their MCFG tables. It maybe that the BIOS is simply buggy and
does support all the busses, but it is more conservative to just fall back
to the old method unless it is certain that memory accesses will work.
driver in Linux 2.6. uscanner was just a simple wrapper around a fifo and
contained no logic, the default interface is now libusb (supported by sane).
Reviewed by: HPS
to the full path of the image that is being executed.
Increase AT_COUNT.
Remove no longer true comment about types used in Linux ELF binaries,
listed types contain FreeBSD-specific entries.
Reviewed by: kan
This code is heavily inspired by Takanori Watanabe's experimental SMP patch
for i386 and large portion was shamelessly cut and pasted from Peter Wemm's
AP boot code.
in FreeBSD 5.x to allow network device drivers to run with Giant
despite the network stack being Giant-free. This significantly
simplifies calls into ioctl() on network interfaces, especially
in the multicast code, as well as eliminates deferred invocation
of interface if_start routines.
Disable the build on device drivers still depending on
IFF_NEEDSGIANT as they no longer compile. They will be removed
in a few weeks if they haven't been made MPSAFE in that time.
Disabled drivers:
if_ar
if_axe
if_aue
if_cdce
if_cue
if_kue
if_ray
if_rue
if_rum
if_sr
if_udav
if_ural
if_zyd
Drivers that were already disabled because of tty changes:
if_ppp
if_sl
Discussed on: arch@
".note.ABI-tag" section.
The search order of a brand is changed, now first of all the
".note.ABI-tag" is looked through.
Move code which fetch osreldate for ELF binary to check_note() handler.
PR: 118473
Approved by: kib (mentor)
system call entry path and i386 IP checksum generation: we now assume
all code is MPSAFE unless explicitly marked otherwise. Remove XXX
Giant comments along similar lines: the code by the comments either
doesn't need or doesn't want Giant (especially the NMI handler).
MFC after: 3 days
ABIs:
- Store the FPU initial control word in the pcb for each thread.
- When first using the FPU, load the initial control word after restoring
the clean state if it is not the standard control word.
- Provide a correct control word for Linux/i386 binaries under
FreeBSD/amd64.
- Adjust the control word returned for fpugetregs()/npxgetregs() when a
thread hasn't used the FPU yet to reflect the real initial control
word for the current ABI.
- The Linux/i386 ABI for FreeBSD/i386 now properly sets the right control
word instead of trashing whatever the current state of the FPU is.
Reviewed by: bde
- Remove the control word parameter to npxinit(). It was always set
to __INITIAL_NPXCW__.
- Remove npx_cleanstate_ready as the cleanstate is always initalized
when it is used.
- Improve the handling of the case when the FPU isn't present. Now
the npx0 device no longer succeeds in its probe so all of npx_attach()
is skipped. Also, we allow this case with SMP (though that shouldn't
actually occur as all i386 systems that support SMP have FPUs) now.
SMP was only an issue back when we had an FPU emulator which was not
per-CPU.
- MFamd64: Clear some of the state in npx_cleanstate rather than leaving
it as garbage.
- MFamd64: When a user thread first uses the FPU, use npx_cleanstate for
the initial FPU state.
Reviewed by: bde
bogus entries have a starting IRQ that is invalid (> 255, so won't fit
into a PCI intline config register). It had the side effect of breaking
MSI by "claiming" several IRQs in the MSI range. Fix this by ignoring such
I/O APICs.
MFC after: 2 weeks
are used by glibc. This silents the message "2.4+ kernel w/o ELF notes?"
from some programs at start, among them are top and pkill.
Do the assignment of the vector entries in elf_linux_fixup()
as it is done in glibc.
Fix some minor style issues.
Submitted by: Marcin Cieslak <saper at SYSTEM PL>
Approved by: kib (mentor)
MFC after: 1 week
different cpu is still assigned to that vector by never clearing idt
entries. This was only provided as a debugging feature and the bugs
are caught by other means.
- Drop the sched lock when rebinding to reassign an interrupt vector
to a new cpu so that pending interrupts have a chance to be delivered
before removing the old vector.
Discussed with: tegge, jhb
the same dma tag. However, it can happen multiple dma tags share the same
bounce zone too, so add a per-bounce zone map counter, and check it instead of
the dma tag map counter, to know if we have to alloc more pages.
Reported by: miwi
Reviewed by: scottl
offset. This is needed for the ehci hardware buffer rings that assume
this behavior.
This is an interim solution, and a more general one is being worked
on. This solution doesn't break anything that doesn't ask for it
directly. The mbuf and uio variants with this flag likely don't work
and haven't been tested.
Universe builds with these changes. I don't have a huge-memory
machine to test these changes with, but will be happy to work with
folks that do and hps if this changes turns out not to be sufficient.
Submitted by: alfred@ from Hans Peter Selasky's original
- leave pmtimer comment that is common to other architectures.
- bring pbio explanation to the block comment relating to other
drivers in the same block.
global NOTES file.
cx(4) driver isn't present in this file, though it could be. However, cx(4)
seems to be more or less dead -- it hasn't been linked to the modules build,
and after TTY-ng transformations it doesn't compile.
Remove it until cx(4) is broken.
- correct format strings
- fill opt_agp.h if AGP_DEBUG is defined
- bring AGP_DEBUG to LINT by mentioning it in NOTES
This should hopefully fix a warning that was...
Found by: Coverity Prevent(tm)
CID: 3676
Tested on: amd64, i386
more irqs as we have more cpus. This is principally useful on systems
with msi devices which may want many irqs per-cpu.
Discussed with: jhb
Sponsored by: Nokia
- add a reference to the config(5) manpage;
- hopefully clarify the format of the 'env FILENAME' directive.
I am putting these notes in sys/${arch}/conf/GENERIC and not
in sys/conf/NOTES because:
1. i386/GENERIC already had reference to a similar option (hints..)
and to documentation (handbook)
2. GENERIC is what most users look at when they have to modify or
create a new kernel config, so having the suggestion there is
more effective.
I am only touching i386 and amd64 because the other GENERIC files
are already out of sync, and I am not sure what is the overall plan.
MFC after: 3 days
to GENERIC configuration files. This brings what's in 8.x in sync
with what is in 7.x, but does not change any current defaults.
Possibly they should now be enabled in head by default?
In the existing code we didn't really enforce that callers hold Giant
before calling userland_sysctl(), even though there is no guarantee it
is safe. Fix this by just placing Giant locks around the call to the oid
handler. This also means we only pick up Giant for a very short period
of time. Maybe we should add MPSAFE flags to sysctl or phase it out all
together.
I've also added SYSCTL_LOCK_ASSERT(). We have to make sure sysctl_root()
and name2oid() are called with the sysctl lock held.
Reviewed by: Jille Timmermans <jille quis cx>
Log:
- merge in latest xenbus from dfr's xenhvm
- fix race condition in xs_read_reply by converting tsleep to mtx_sleep
Log:
unmask evtchn in bind_{virq, ipi}_to_irq
Log:
- remove code for handling case of not being able to sleep
- eliminate tsleep - make sleeps atomic
time it is marked for user space callchain capture in the NMI
handler and the time the callchain capture callback runs.
- Improve code and control flow clarity by invoking hwpmc(4)'s user
space callchain capture callback directly from low-level code.
Reviewed by: jhb (kern/subr_trap.c)
Testing (various patch revisions): gnn,
Fabien Thomas <fabien dot thomas at netasq dot com>,
Artem Belevich <artemb at gmail dot com>
memory barriers on i386. It works as a serialization instruction on
all IA32 CPUs.
Alternative solution of using {s,l,}fence requires run-time checking
of the presense of the corresponding SSE or SSE2 extensions, and
possible boot-time patching of the kernel text.
Suggested by: many
Sgtty is a programming interface that has been replaced by termios over
the years. In June we already removed <sgtty.h>, which exposes the
ioctl()'s that are implemented by this interface. The importance of this
flag is overrated right now.
module; the ath module now brings in the hal support. Kernel
config files are almost backwards compatible; supplying
device ath_hal
gives you the same chip support that the binary hal did but you
must also include
options AH_SUPPORT_AR5416
to enable the extended format descriptors used by 11n parts.
It is now possible to control the chip support included in a
build by specifying exactly which chips are to be supported
in the config file; consult ath_hal(4) for information.
Volume 3B: System Programming Guide, Part 2", CPUs with family 0x6 and model
above or 0xE and CPUs with family 0xF and model above or 0x3 have invariant
TSC.
Change types used in the linux' struct msghdr and struct cmsghdr
definitions to the properly-sized architecture-specific types.
Move ancillary data handler from linux_sendit() to linux_sendmsg().
Submitted by: dchagin
and Core Duo), models 0xF (Core2), model 0x17 (Core2Extreme) and
model 0x1C (Atom).
In these CPUs, the actual numbers, kinds and widths of PMCs present
need to queried at run time. Support for specific "architectural"
events also needs to be queried at run time.
Model 0xE CPUs support programmable PMCs, subsequent CPUs
additionally support "fixed-function" counters.
- Use event names that are close to vendor documentation, taking in
account that:
- events with identical semantics on two or more CPUs in this family
can have differing names in vendor documentation,
- identical vendor event names may map to differing events across
CPUs,
- each type of CPU supports a different subset of measurable
events.
Fixed-function and programmable counters both use the same vendor
names for events. The use of a class name prefix ("iaf-" or
"iap-" respectively) permits these to be distinguished.
- In libpmc, refactor pmc_name_of_event() into a public interface
and an internal helper function, for use by log handling code.
- Minor code tweaks: staticize a global, freshen a few comments.
Tested by: gnn
throttle. My SMP kernel hangs when one of those is selected by
powerd. Errata AA21 here:
ftp://download.intel.com/design/PentiumXE/specupdt/31030717.pdf
MFC after: 2 weeks
of the ABI of the currently executing image. Change some places to test
the flags instead of explicit comparing with address of known sysentvec
structures to determine ABI features.
Discussed with: dchagin, imp, jhb, peter
and ifnet functions
- add memory barriers to <machine/atomic.h>
- update drivers to only conditionally define their own
- add lockless producer / consumer ring buffer
- remove ring buffer implementation from cxgb and update its callers
- add if_transmit(struct ifnet *ifp, struct mbuf *m) to ifnet to
allow drivers to efficiently manage multiple hardware queues
(i.e. not serialize all packets through one ifq)
- expose if_qflush to allow drivers to flush any driver managed queues
This work was supported by Bitgravity Inc. and Chelsio Inc.
controller. The controller is also known as L1E(AR8121) and
L2E(AR8113/AR8114). Unlike its predecessor Attansic L1,
AR8121/AR8113/AR8114 uses completely different Rx logic such that
it requires separate driver. Datasheet for AR81xx is not available
to open source driver writers but it shares large part of Tx and
PHY logic of L1. I still don't understand some part of register
meaning and some MAC statistics counters but the driver seems to
have no critical issues for performance and stability.
The AR81xx requires copy operation to pass received frames to upper
stack such that ale(4) consumes a lot of CPU cycles than that of
other controller. A couple of silicon bugs also adds more CPU
cycles to address the known hardware bug. However, if you have fast
CPU you can still saturate the link.
Currently ale(4) supports the following hardware features.
- MSI.
- TCP Segmentation offload.
- Hardware VLAN tag insertion/stripping with checksum offload.
- Tx TCP/UDP checksum offload and Rx IP/TCP/UDP checksum offload.
- Tx/Rx interrupt moderation.
- Hardware statistics counters.
- Jumbo frame.
- WOL.
AR81xx PCIe ethernet controllers are mainly found on ASUS EeePC or
P5Q series of ASUS motherboards. Special thanks to Jeremy Chadwick
who sent the hardware to me. Without his donation writing a driver
for AR81xx would never have been possible. Big thanks to all people
who reported feedback or tested patches.
HW donated by: koitsu
Tested by: bsam, Joao Barros <joao.barros <> gmail DOT com >
Jan Henrik Sylvester <me <> janh DOT de >
Ivan Brawley < ivan <> brawley DOT id DOT au >,
CURRENT ML
