Currently, Application Processors (non-boot CPUs) are started by
MD code at SI_SUB_CPU, but they are kept waiting in a "pen" until
SI_SUB_SMP at which point they are released to run kernel threads.
SI_SUB_SMP is one of the last SYSINIT levels, so APs don't enter
the scheduler and start running threads until fairly late in the
boot.
This change moves SI_SUB_SMP up to just before software interrupt
threads are created allowing the APs to start executing kernel
threads much sooner (before any devices are probed). This allows
several initialization routines that need to perform initialization
on all CPUs to now perform that initialization in one step rather
than having to defer the AP initialization to a second SYSINIT run
at SI_SUB_SMP. It also permits all CPUs to be available for
handling interrupts before any devices are probed.
This last feature fixes a problem on with interrupt vector exhaustion.
Specifically, in the old model all device interrupts were routed
onto the boot CPU during boot. Later after the APs were released at
SI_SUB_SMP, interrupts were redistributed across all CPUs.
However, several drivers for multiqueue hardware allocate N interrupts
per CPU in the system. In a system with many CPUs, just a few drivers
doing this could exhaust the available pool of interrupt vectors on
the boot CPU as each driver was allocating N * mp_ncpu vectors on the
boot CPU. Now, drivers will allocate interrupts on their desired CPUs
during boot meaning that only N interrupts are allocated from the boot
CPU instead of N * mp_ncpu.
Some other bits of code can also be simplified as smp_started is
now true much earlier and will now always be true for these bits of
code. This removes the need to treat the single-CPU boot environment
as a special case.
As a transition aid, the new behavior is available under a new kernel
option (EARLY_AP_STARTUP). This will allow the option to be turned off
if need be during initial testing. I plan to enable this on x86 by
default in a followup commit in the next few days and to have all
platforms moved over before 11.0. Once the transition is complete,
the option will be removed along with the !EARLY_AP_STARTUP code.
These changes have only been tested on x86. Other platform maintainers
are encouraged to port their architectures over as well. The main
things to check for are any uses of smp_started in MD code that can be
simplified and SI_SUB_SMP SYSINITs in MD code that can be removed in
the EARLY_AP_STARTUP case (e.g. the interrupt shuffling).
PR: kern/199321
Reviewed by: markj, gnn, kib
Sponsored by: Netflix
bus_get_cpus() returns a specified set of CPUs for a device. It accepts
an enum for the second parameter that indicates the type of cpuset to
request. Currently two valus are supported:
- LOCAL_CPUS (on x86 this returns all the CPUs in the package closest to
the device when DEVICE_NUMA is enabled)
- INTR_CPUS (like LOCAL_CPUS but only returns 1 SMT thread for each core)
For systems that do not support NUMA (or if it is not enabled in the kernel
config), LOCAL_CPUS fails with EINVAL. INTR_CPUS is mapped to 'all_cpus'
by default. The idea is that INTR_CPUS should always return a valid set.
Device drivers which want to use per-CPU interrupts should start using
INTR_CPUS instead of simply assigning interrupts to all available CPUs.
In the future we may wish to add tunables to control the policy of
INTR_CPUS (e.g. should it be local-only or global, should it ignore
SMT threads or not).
The x86 nexus driver exposes the internal set of interrupt CPUs from the
the x86 interrupt code via INTR_CPUS.
The ACPI bus driver and PCI bridge drivers use _PXM to return a suitable
LOCAL_CPUS set when _PXM exists and DEVICE_NUMA is enabled. They also and
the global INTR_CPUS set from the nexus driver with the per-domain set from
_PXM to generate a local INTR_CPUS set for child devices.
Compared to the r298933, this version uses 'struct _cpuset' in
<sys/bus.h> instead of 'cpuset_t' to avoid requiring <sys/param.h>
(<sys/_cpuset.h> still requires <sys/param.h> for MAXCPU even though
<sys/_bitset.h> does not after recent changes).
bus_get_cpus() returns a specified set of CPUs for a device. It accepts
an enum for the second parameter that indicates the type of cpuset to
request. Currently two valus are supported:
- LOCAL_CPUS (on x86 this returns all the CPUs in the package closest to
the device when DEVICE_NUMA is enabled)
- INTR_CPUS (like LOCAL_CPUS but only returns 1 SMT thread for each core)
For systems that do not support NUMA (or if it is not enabled in the kernel
config), LOCAL_CPUS fails with EINVAL. INTR_CPUS is mapped to 'all_cpus'
by default. The idea is that INTR_CPUS should always return a valid set.
Device drivers which want to use per-CPU interrupts should start using
INTR_CPUS instead of simply assigning interrupts to all available CPUs.
In the future we may wish to add tunables to control the policy of
INTR_CPUS (e.g. should it be local-only or global, should it ignore
SMT threads or not).
The x86 nexus driver exposes the internal set of interrupt CPUs from the
the x86 interrupt code via INTR_CPUS.
The ACPI bus driver and PCI bridge drivers use _PXM to return a suitable
LOCAL_CPUS set when _PXM exists and DEVICE_NUMA is enabled. They also and
the global INTR_CPUS set from the nexus driver with the per-domain set from
_PXM to generate a local INTR_CPUS set for child devices.
Reviewed by: wblock (manpage)
Differential Revision: https://reviews.freebsd.org/D5519
Both of the callers were expecting the input cap_set to be modified.
This fixes them to request cap_set to be updated with the returned buffer.
Reviewed by: jkim
Differential Revision: https://reviews.freebsd.org/D6040
This wrapper does not translate errors in the first word to ACPI
error status returns. Use this wrapper in the acpi_cpu(4) driver in
place of the existing _OSC code. While here, fix a bug where the wrong
count of words was passed when invoking _OSC.
Reviewed by: jkim
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D6022
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.
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D5782
On some architectures, u_long isn't large enough for resource definitions.
Particularly, powerpc and arm allow 36-bit (or larger) physical addresses, but
type `long' is only 32-bit. This extends rman's resources to uintmax_t. With
this change, any resource can feasibly be placed anywhere in physical memory
(within the constraints of the driver).
Why uintmax_t and not something machine dependent, or uint64_t? Though it's
possible for uintmax_t to grow, it's highly unlikely it will become 128-bit on
32-bit architectures. 64-bit architectures should have plenty of RAM to absorb
the increase on resource sizes if and when this occurs, and the number of
resources on memory-constrained systems should be sufficiently small as to not
pose a drastic overhead. That being said, uintmax_t was chosen for source
clarity. If it's specified as uint64_t, all printf()-like calls would either
need casts to uintmax_t, or be littered with PRI*64 macros. Casts to uintmax_t
aren't horrible, but it would also bake into the API for
resource_list_print_type() either a hidden assumption that entries get cast to
uintmax_t for printing, or these calls would need the PRI*64 macros. Since
source code is meant to be read more often than written, I chose the clearest
path of simply using uintmax_t.
Tested on a PowerPC p5020-based board, which places all device resources in
0xfxxxxxxxx, and has 8GB RAM.
Regression tested on qemu-system-i386
Regression tested on qemu-system-mips (malta profile)
Tested PAE and devinfo on virtualbox (live CD)
Special thanks to bz for his testing on ARM.
Reviewed By: bz, jhb (previous)
Relnotes: Yes
Sponsored by: Alex Perez/Inertial Computing
Differential Revision: https://reviews.freebsd.org/D4544
This simplifies checking for default resource range for bus_alloc_resource(),
and improves readability.
This is part of, and related to, the migration of rman_res_t from u_long to
uintmax_t.
Discussed with: jhb
Suggested by: marcel
Summary:
Migrate to using the semi-opaque type rman_res_t to specify rman resources. For
now, this is still compatible with u_long.
This is step one in migrating rman to use uintmax_t for resources instead of
u_long.
Going forward, this could feasibly be used to specify architecture-specific
definitions of resource ranges, rather than baking a specific integer type into
the API.
This change has been broken out to facilitate MFC'ing drivers back to 10 without
breaking ABI.
Reviewed By: jhb
Sponsored by: Alex Perez/Inertial Computing
Differential Revision: https://reviews.freebsd.org/D5075
to shut down; close laptop lid" scenario which otherwise tended to end
with a laptop overheating or the battery dying.
The implementation uses a new sysctl, kern.suspend_blocked; init(8) sets
this while rc.suspend runs, and the ACPI sleep code ignores requests while
the sysctl is set.
Discussed on: freebsd-acpi (35 emails)
MFC after: 1 week
years for head. However, it is continuously misused as the mpsafe argument
for callout_init(9). Deprecate the flag and clean up callout_init() calls
to make them more consistent.
Differential Revision: https://reviews.freebsd.org/D2613
Reviewed by: jhb
MFC after: 2 weeks
its use in upcoming code.
This is inspired by something in jhb's NUMA IRQ allocation patchset.
However, the tricky bit here is that the PXM lookup for a node may
fail, requiring a lookup on the parent node. So if it doesn't
exist, don't fail - just go up to the parent. Only error out of the
lookup is the ACPI lookup returns an error.
Sponsored by: Norse Corp, Inc.
under bootverbose. Every example I've seen to date has been due to
an ACPI system resource device reserving a range that overlaps with
system memory (which ram0 attempts to reserve) or a local or I/O APIC
(which apic0 attempts to reserve). These are always harmless but look
scary to users.
MFC after: 1 week
allows the user to request administrative changes to individual devices
such as attach or detaching drivers or disabling and re-enabling devices.
- Add a new /dev/devctl2 character device which uses ioctls for device
requests. The ioctls use a common 'struct devreq' which is somewhat
similar to 'struct ifreq'.
- The ioctls identify the device to operate on via a string. This
string can either by the device's name, or it can be a bus-specific
address. (For unattached devices, a bus address is the only way to
locate a device.) Bus drivers register an eventhandler to claim
unrecognized device names that the driver recognizes as a valid address.
Two buses currently support addresses: ACPI recognizes any device
in the ACPI namespace via its full path starting with "\" and
the PCI bus driver recognizes an address specification of
'pci[<domain>:]<bus>:<slot>:<func>' (identical to the PCI selector
strings supported by pciconf).
- To make it easier to cut and paste, change the PnP location string
in the PCI bus driver to output a full PCI selector string rather
than 'slot=<slot> function=<func>'.
- Add a devctl(3) interface in libdevctl which provides a wrapper around
the ioctls and is the preferred interface for other userland code.
- Add a devctl(8) program which is a simple wrapper around the requests
supported by devctl(3).
- Add a device_is_suspended() function to check DF_SUSPENDED.
- Add a resource_unset_value() function that can be used to remove a
hint from the kernel environment. This is used to clear a
hint.<driver>.<unit>.disabled hint when re-enabling a boot-time
disabled device.
Reviewed by: imp (parts)
Requested by: imp (changing PCI location string)
Relnotes: yes
Also, split power_suspend into power_suspend and power_suspend_early.
power_suspend_early is called before the userland is frozen.
power_suspend is called after the userland is frozen.
Currently only VT switching is hooked to power_suspend_early.
This is needed because switching away from X server requires its
cooperation, so obviously X server must not be frozen when that happens.
Freezing userland during ACPI suspend is useful because not all drivers
correctly handle suspension concurrent with other activity. This is
especially applicable to drivers ported from other operating systems
that suspend all software activity between placing drivers and hardware
into suspended state.
In particular drm2/radeon (radeonkms) depends on the described
procedure. The driver does not have any internal synchronization
between suspension activities and processing of userland requests.
Many thanks to kib for the code that allows to freeze and thaw all
userland threads.
Note that ideally we also need to park / inhibit (non-special) kernel
threads as well to ensure that they do not call into drivers.
MFC after: 17 days
state said device should go into.
This was a snafu introduced in the ACPI/PCI awareness separation.
When putting a device into a power state, the bus (and thus firmware,
eg ACPI) should be asked before hand to check whether the device
can indeed go into that power state.
There's a set of nodes in ACPI under each device - the _SxD nodes - which
state which ACPI power state to put the device into when the system is
going into power save state 'x'. So when going into S3, the existence
of an _S3D node would override whatever the system was trying to do.
By default the PCI code wants to put devices into D3 before suspending.
I have a laptop here (Asus Zenbook - check the PR) whose EHCI controller
really wants to be in D2 during suspend, not D3. So if we put it into
D3 and then try to enter S3, everything hangs. The device itself
can go into D3 - it just can't be there when the call to ACPI to enter
S3 occurs. The PCI patch fixes this.
jkim@ noticed that the same is needed for the ACPI child device
enumeration.
Thankyou to Matt Dillon (the programmer, not the actor) for buying me
this particular laptop so I could debug the issues with the Atheros
AR9485 that is in it. It's his fault that I ended up with this
laptop and was sufficiently annoyed by the lack of USB suspend
to go down this rabbit hole.
Tested:
* Thinkpad T400
* Thinkpad X230
* Thinkpad T42
* Thinkpad T60
* Asus Zenbook (see PR)
* Asus EEEPC 701
* Asus EEEPC 1001PX
TODO:
* Figure out what we should do about devices we unload drivers for
that want to be in a specific state when entering S3 / S4 -
the "put devices into D3 if they're not bound to a driver" option
may also mess with things.
PR: kern/194884
Reviewed by: jhb, jkim
MFC after: 1 week
Relnotes: yes
Sponsored by: Matt Dillon <dillon@apollo.backplane.com> (hardware)
directly accessed. Although this will work on some platforms, it can
throw an exception if the pointer is invalid and then panic the kernel.
Add a missing SYSCTL_IN() of "SCTP_BASE_STATS" structure.
MFC after: 3 days
Sponsored by: Mellanox Technologies
in userland rename in-kernel getenv()/setenv() to kern_setenv()/kern_getenv().
This fixes a namespace collision with libc symbols.
Submitted by: kmacy
Tested by: make universe
* Add a bus_if.m method - get_domain() - returning the VM domain or
ENOENT if the device isn't in a VM domain;
* Add bus methods to print out the domain of the device if appropriate;
* Add code in srat.c to save the PXM -> VM domain mapping that's done and
expose a function to translate VM domain -> PXM;
* Add ACPI and ACPI PCI methods to check if the bus has a _PXM attribute
and if so map it to the VM domain;
* (.. yes, this works recursively.)
* Have the pci bus glue print out the device VM domain if present.
Note: this is just the plumbing to start enumerating information -
it doesn't at all modify behaviour.
Differential Revision: D906
Reviewed by: jhb
Sponsored by: Norse Corp
resume that is a superset of a pcb. Move the FPU state out of the pcb and
into this new structure. As part of this, move the FPU resume code on
amd64 into a C function. This allows resumectx() to still operate only on
a pcb and more closely mirrors the i386 code.
Reviewed by: kib (earlier version)
PCI root bridges except for the one known-valid case on x86 where bridges
claim the I/O port registers used for PCI config space access.
Tested by: Hilko Meyer <hilko.meyer@gmx.de>
MFC after: 1 week
the memory ranges that they decode for downstream devices rather than
creating ResourceProducer range resource entries. The result is that
we allocate the full range to the PCI root bridge device causing
allocations in child devices to all fail.
As a workaround, ignore any standard memory resources on a PCI root
bridge device. It is normal for a PCI root bridge to allocate an I/O
resource for the I/O ports used for PCI config access, but I have not
seen any PCI root bridges that legitimately allocate a memory resource.
Reviewed by: jkim
MFC after: 1 week
resist easy conversion since they implement a great deal of their attach
logic inside probe(). Some of this could be fixed by moving it to attach(),
but some requires something more subtle than BUS_PROBE_NOWILDCARD.
that uses non-ISA IRQs but use a plain IRQ resource in _CRS. However,
a non-ISA IRQ can't fit into a plain IRQ resource. If we encounter a
link like this, build the resource buffer from _PRS instead of _CRS.
- Set the correct size of the end tag in a resource buffer.
Tested by: Benjamin Lee <ben@b1c1l1.com>
MFC after: 2 weeks
This hack is picked up from Linux, which claims that it follows
Windows behavior.
PR: amd64/174409
Tested by: Sergey V. Dyatko <sergey.dyatko@gmail.com>,
KAHO Toshikazu <kaho@elam.kais.kyoto-u.ac.jp>,
Slawa Olhovchenkov <slw@zxy.spb.ru>
MFC after: 13 days
