Do not try to change attributes for DMAP when working on a mapping
which is not covered by the DMAP. This was reported on real system
where a BAR of a device (NTB) was mapped outside the PCI window.
Reported and tested by: mav
Reviewed by: jhb, mav
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Differential revision: https://reviews.freebsd.org/D6668
The only current purpose of the pvh lock was explained there
On Wed, Jan 09, 2013 at 11:46:13PM -0600, Alan Cox wrote:
> Let me lay out one example for you in detail. Suppose that we have
> three processors and two of these processors are actively using the same
> pmap. Now, one of the two processors sharing the pmap performs a
> pmap_remove(). Suppose that one of the removed mappings is to a
> physical page P. Moreover, suppose that the other processor sharing
> that pmap has this mapping cached with write access in its TLB. Here's
> where the trouble might begin. As you might expect, the processor
> performing the pmap_remove() will acquire the fine-grained lock on the
> PV list for page P before destroying the mapping to page P. Moreover,
> this processor will ensure that the vm_page's dirty field is updated
> before releasing that PV list lock. However, the TLB shootdown for this
> mapping may not be initiated until after the PV list lock is released.
> The processor performing the pmap_remove() is not problematic, because
> the code being executed by that processor won't presume that the mapping
> is destroyed until the TLB shootdown has completed and pmap_remove() has
> returned. However, the other processor sharing the pmap could be
> problematic. Specifically, suppose that the third processor is
> executing the page daemon and concurrently trying to reclaim page P.
> This processor performs a pmap_remove_all() on page P in preparation for
> reclaiming the page. At this instant, the PV list for page P may
> already be empty but our second processor still has a stale TLB entry
> mapping page P. So, changes might still occur to the page after the
> page daemon believes that all mappings have been destroyed. (If the PV
> entry had still existed, then the pmap lock would have ensured that the
> TLB shootdown completed before the pmap_remove_all() finished.) Note,
> however, the page daemon will know that the page is dirty. It can't
> possibly mistake a dirty page for a clean one. However, without the
> current pvh global locking, I don't think anything is stopping the page
> daemon from starting the laundering process before the TLB shootdown has
> completed.
>
> I believe that a similar example could be constructed with a clean page
> P' and a stale read-only TLB entry. In this case, the page P' could be
> "cached" in the cache/free queues and recycled before the stale TLB
> entry is flushed.
TLBs for addresses with updated PTEs are always flushed before pmap
lock is unlocked. On the other hand, amd64 pmap code does not always
flushes TLBs before PV list locks are unlocked, if previously PTEs
were cleared and PV entries removed.
To handle the situations where a thread might notice empty PV list but
third thread still having access to the page due to TLB invalidation
not finished yet, introduce delayed invalidation. Comparing with the
pvh_global_lock, DI does not block entered thread when
pmap_remove_all() or pmap_remove_write() (callers of
pmap_delayed_invl_wait()) are executed in parallel. But _invl_wait()
callers are blocked until all previously noted DI blocks are leaved,
thus ensuring that neccessary TLB invalidations were performed before
returning from pmap_remove_all() or pmap_remove_write().
See comments for detailed description of the mechanism, and also for
the explanations why several pmap methods, most important
pmap_enter(), do not need DI protection.
Reviewed by: alc, jhb (turnstile KPI usage)
Tested by: pho (previous version)
Sponsored by: The FreeBSD Foundation
Differential revision: https://reviews.freebsd.org/D5747
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).
PCI-express HotPlug support is implemented via bits in the slot
registers of the PCI-express capability of the downstream port along
with an interrupt that triggers when bits in the slot status register
change.
This is implemented for FreeBSD by adding HotPlug support to the
PCI-PCI bridge driver which attaches to the virtual PCI-PCI bridges
representing downstream ports on HotPlug slots. The PCI-PCI bridge
driver registers an interrupt handler to receive HotPlug events. It
also uses the slot registers to determine the current HotPlug state
and drive an internal HotPlug state machine. For simplicty of
implementation, the PCI-PCI bridge device detaches and deletes the
child PCI device when a card is removed from a slot and creates and
attaches a PCI child device when a card is inserted into the slot.
The PCI-PCI bridge driver provides a bus_child_present which claims
that child devices are present on HotPlug-capable slots only when a
card is inserted. Rather than requiring a timeout in the RC for
config accesses to not-present children, the pcib_read/write_config
methods fail all requests when a card is not present (or not yet
ready).
These changes include support for various optional HotPlug
capabilities such as a power controller, mechanical latch,
electro-mechanical interlock, indicators, and an attention button.
It also includes support for devices which require waiting for
command completion events before initiating a subsequent HotPlug
command. However, it has only been tested on ExpressCard systems
which support surprise removal and have none of these optional
capabilities.
PCI-express HotPlug support is conditional on the PCI_HP option
which is enabled by default on arm64, x86, and powerpc.
Reviewed by: adrian, imp, vangyzen (older versions)
Relnotes: yes
Differential Revision: https://reviews.freebsd.org/D6136
call pmap_invalidate_page() even though they are not destroying a leaf-
level page table entry.
Eliminate some bogus white-space characters in a comment.
Reviewed by: kib
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
Summary:
The Initial Local APIC ID is returned by CPUID function 1 (in EBX).
On AMD Family 10h systems the way that ID is built is controlled by
an MSR bit (InitApicIdCpuIdLo). BKDG instructs BIOS to set it in a
certain way, but a BIOS can be buggy. In that case the ID can confuse
tools that use it, e.g. hwloc.
For example, on a system that I own real Local APIC IDs are configured
as 0, 1, 2, 3, but IDs reported via CPUID.1 are 0, 0x40, 0x80, 0xc0.
See: https://github.com/open-mpi/hwloc/issues/183
Reviewed by: kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D6060
rounddown2 tends to produce longer lines than the original code
and when the code has a high indentation level it was not really
advantageous to do the replacement.
This tries to strike a balance between readability using the macros
and flexibility of having the expressions, so not everything is
converted.
(And 4Kn minidump support, but only for amd64.)
Make sure all I/O to the dump device is of the native sector size. To
that end, we keep a native sector sized buffer associated with dump
devices (di->blockbuf) and use it to pad smaller objects as needed (e.g.
kerneldumpheader).
Add dump_write_pad() as a convenience API to dump smaller objects with
zero padding. (Rather than pull in NPM leftpad, we wrote our own.)
Savecore(1) has been updated to deal with these dumps. The format for
512-byte sector dumps should remain backwards compatible.
Minidumps for other architectures are left as an exercise for the
reader.
PR: 194279
Submitted by: ambrisko@
Reviewed by: cem (earlier version), rpokala
Tested by: rpokala (4Kn/512 except 512 fulldump), cem (512 fulldump)
Relnotes: yes
Sponsored by: EMC / Isilon Storage Division
Differential Revision: https://reviews.freebsd.org/D5848
Submitted by: Jun Su <junsu microsoft com>
Reviewed by: jhb, kib, sephe
Sponsored by: Microsoft OSTC
Differential Revision: https://reviews.freebsd.org/D5910
doreti provides the common code path for returning from interrupt
andlers on x86. Exposing doreti as a global symbol allows kernel
modules to include low-level interrupt handlers instead of requiring
all low-level handlers to be statically compiled into the kernel.
Submitted by: Howard Su <howard0su@gmail.com>
Reviewed by: kib
Some BIOSes disable AMD Topology extension on AMD Family 15h notebook
processors. We re-enable the extension, so that we can properly discover
core and cache topology. Linux seems to do the same.
Reported by: Johannes Dieterich <dieterich.joh@gmail.com>
Reviewed by: jhb, kib
Tested by: Johannes Dieterich <dieterich.joh@gmail.com>
(earlier version)
MFC after: 3 weeks
Differential Revision: https://reviews.freebsd.org/D5883
DTrace-related exceptions in userland code are handled elsewhere.
One practical problem was a crash in dtrace_invop_start() when saved
%rsp pointed to a virtual address that was not backed.
i386 code already ignored userland exceptions.
Reviewed by: markj, kib
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D5906
We're currently seeing how hard it would be to run CloudABI binaries on
operating systems cannot be modified easily (Windows, Mac OS X). The
idea is that we want to just run them without any sandboxing. Now
that CloudABI executables are PIE, this is already a bit easier, but TLS
is still problematic:
- CloudABI executables want to write to the %fs, which typically
requires extra system calls by the emulator every time it needs to
switch between CloudABI's and its own TLS.
- If CloudABI executables overwrite the %fs base unconditionally, it
also becomes harder for the emulator to store a backup of the old
value of %fs. To solve this, let's no longer overwrite %fs, but just
%fs:0.
As CloudABI's C library does not use a TCB, this space can now be used
by an emulator to keep track of its internal state. The executable can
now safely overwrite %fs:0, as long as it makes sure that the TCB is
copied over to the new TLS area.
Ensure that there is an initial TLS area set up when the process starts,
only containing a bogus TCB. We don't really care about its contents on
FreeBSD.
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D5836
kern.features.linux: 1 meaning linux 32 bits binaries are supported
kern.features.linux64: 1 meaning linux 64 bits binaries are supported
The goal here is to help 3rd party applications (including ports) to determine
if the host do support linux emulation
Reviewed by: dchagin
MFC after: 1 week
Relnotes: yes
Differential Revision: D5830
- Set BI_CAN_EXEC_DYN, so we can execute ET_DYN ELF files in addition to
regular ET_EXECs.
- Provide an AT_BASE entry in the auxiliary vector, so the executable
knows at which address it got loaded and can apply relocations.
Simplify and unify placeholder type definitions.
Reviewed by: jhb
Sponsored by: The FreeBSD Foundation
Differential revision: https://reviews.freebsd.org/D5771
This moves the enabling of interrupts slightly earlier (the old location
was still before devices were enumerated and probed) and does it in the
interrupt code (rather than in the device configuration code). This
also avoids tripping over an assertion on the first TLB shootdown with
earlier AP startup.
Reviewed by: kib
Sponsored by: Netflix
Differential Revision: https://reviews.freebsd.org/D5710
during argument validity verification, unbound zero'ing of the process LDT
and adjacent memory can be initiated from usermode.
Submitted by: CORE Security
Patch by: kib
Security: SA-16:15
non-multiple of 64 bytes. Thereafter, the user state save area is
misaligned, which triggers assertion in the debugging kernels, or
segmentation violation on accesses for non-debugging configs.
Force the desired alignment of the user save area as the fix
(workaround is to disable bit 9 in the hw.xsave_mask loader tunable).
This correction is required for booting on the upcoming Intel' Purley
platform.
Reported and tested by: "Pieper, Jeffrey E" <jeffrey.e.pieper@intel.com>,
jimharris
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
- Advertise the word size for CloudABI ABIs via the SV_LP64 flag. All of
the other ABIs include either SV_ILP32 or SV_LP64.
- Fix kdump to not assume a 32-bit ABI if the ABI flags field is non-zero
but SV_LP64 isn't set. Instead, only assume a 32-bit ABI if SV_ILP32 is
set and fallback to the unknown value of "00" if neither SV_LP64 nor
SV_ILP32 is set.
Reviewed by: kib, ed
Differential Revision: https://reviews.freebsd.org/D5560
