Svatopluk Kraus [Tue, 24 Nov 2015 13:57:41 +0000 (13:57 +0000)]
Flush all kernel mappings from TLB(s) in time when they are cleared.
Replace tlb_flush_local() by tlb_flush() as even not global mappings
could be fetched to TLB(s) on other cores by speculative table walk.
From OS point of view, it was not a problem as either such mappings
were not used anymore or they were flushed from TLB(s) when reused.
However, from hardware point of view, it was a problem. Not flushed
mappings could be a target for speculative reads or prefetches (which
might be quite aggresive on ARM cores). As speculative read can fill
cacheline, it can cause a real problem, when physical page is reused,
but mapped with different memory attributes.
Anyhow, it's good to have only valid mappings in TLB(s).
Andrew Turner [Tue, 24 Nov 2015 11:01:43 +0000 (11:01 +0000)]
Add support for moving the DMAP range. This is needed as some AMD SoCs
place physical memory at an address outside the old DMAP range. This is an
issue as we rely on being able to move from PA -> VA using this range.
Obtained from: Patrick Wildt <patrick@bitrig.org> (earlier version)
Sponsored by: ABT Systems Ltd
Differential Revision: https://reviews.freebsd.org/D3885
Rework the vnode cache recycling to meet free and unused vnodes
targets. See the comment above wantfreevnodes variable for the
description of the algorithm.
The vfs.vlru_alloc_cache_src sysctl is removed. New code frees
namecache sources as the last chance to satisfy the highest watermark,
instead of selecting the source vnodes randomly. This provides good
enough behaviour to keep vn_fullpath() working in most situations.
The filesystem layout with deep trees, where the removed knob was
required, is thus handled automatically.
On PowerPC 64bit, the linux-compat mb() definition is implemented with
lwsync instruction, which does not provide Store/Load barrier. Fix
this by using "full" sync barrier for mb().
atomic_store_rel() does not need full barrier, change mb() call there
to the lwsync instruction if not hitting the known CPU erratas
(i.e. on 32bit). Provide powerpc_lwsync() helper to isolate the
lwsync/sync compile time selection, and use it in atomic_store_rel()
and several other places which duplicate the code.
Noted by: alc
Reviewed and tested by: nwhitehorn
Sponsored by: The FreeBSD Foundation
Bryan Drewery [Tue, 24 Nov 2015 00:49:29 +0000 (00:49 +0000)]
Fix ld not respecting --sysroot.
ld(1) uses the /usr/libdata/ldscripts when linking. These scripts add in the
default search paths of /lib and /usr/lib via 'SEARCH_DIR("DIR")'. These
need to be prefixed by '=' so that the --sysroot flag is respected. This
is not a problem with buildworld since the TOOLS_PREFIX is baked into the
cross-ld. However it is a problem when trying to use ld(1) with --sysroot
anywhere else as it ends up still reading /lib and /usr/lib despite --sysroot.
The default --sysroot (TARGET_SYSTEM_ROOT) is '/' for /usr/bin/ld.
I found this while building with META MODE with uses only --sysroot with
/usr/bin/ld, and found that libraries that I had not built in its sysroot
directory were leaking in. This didn't happen with ports binutils either. This
would also impact external compiler support.
John Baldwin [Mon, 23 Nov 2015 23:48:07 +0000 (23:48 +0000)]
Add a new -B flag for use with list mode (-l) that lists details about
bridges. Currently this includes information about what resources a
bridge decodes on the upstream side for use by downstream devices including
bus numbers, I/O port resources, and memory resources. Windows and bus
ranges are enumerated for both PCI-PCI bridges and PCI-CardBus bridges.
To simplify the implementation, all enumeration is done by reading the
appropriate config space registers directly rather than querying the
bridge driver in the kernel via new ioctls. This does result in a few
limitations.
First, an unimplemented window in a PCI-PCI bridge cannot be accurately
detected as accurate detection requires writing to the window base
register. That is not safe for pciconf(8). Instead, this assumes that
any window where both the base and limit read as all zeroes is
unimplemented.
Second, the PCI-PCI bridge driver in a tree has a few quirks for
PCI-PCI bridges that use subtractive decoding but do not indicate that
via the progif config register. The list of quirks is duplicated in
pciconf's source.
Add destroy_object callback to object rewriting framework.
It is called when last reference to named object is going to be released
and allows to do additional cleanup for implementation of named objects.
Nathan Whitehorn [Mon, 23 Nov 2015 17:07:51 +0000 (17:07 +0000)]
Provide support for userland binaries using the new ELFv2 ABI. This is a
new, simplified, ELF ABI that avoids some of the stranger aspects of the
existing 64-bit PowerPC ABI (function descriptors, in particular). Actually
generating such executables requires a new version of binutils and a newer
compiler (either GCC or clang) than GCC 4.2.1.
Enji Cooper [Mon, 23 Nov 2015 09:15:20 +0000 (09:15 +0000)]
Fix up convert.c generation
- Use a temporary file for convert.c to reduce likelihood of an interrupted
build resulting in bad code being written to convert.c
- Truncate the file instead of appending to it to ensure that the file being
touched will not result in duplicate declarations/definitions from
kern_acct.c if/when kern_acct.c changes.
Split kerne timekeep ABI structure vdso_sv_tk out of the struct
sysentvec. This allows the timekeep data to be shared between similar
ABIs which cannot share sysentvec.
Make the timekeep_push_vdso() tick callback to the timekeep structures
instead of sysentvecs. If several sysentvec share the vdso_sv_tk
structure, we would update the userspace data several times on each
tick, without the change.
Only allocate vdso_sv_tk in the exec_sysvec_init() sysinit when
sysentvec is marked with the new SV_TIMEKEEP flag. This saves
allocation and update of unneeded vdso_sv_tk for ABIs which do not
provide userspace gettimeofday yet, which are PowerPCs arches right
now.
Make vdso_sv_tk allocator public, namely split out and export
alloc_sv_tk() and alloc_sv_tk_compat32(). ABIs which share timekeep
data now can allocate it manually and share as appropriate.
Requested by: nwhitehorn
Tested by: nwhitehorn, pho
Sponsored by: The FreeBSD Foundation
MFC after: 2 weeks
Ian Lepore [Sun, 22 Nov 2015 20:38:50 +0000 (20:38 +0000)]
Print more detailed info about the disk and partition chosen for booting.
No behavioral changes, just cosmetics.
A partition number of zero is not a wildcard, it's the 'a' partition in
a BSD slice, so don't print it as "<auto>". (Only slices are 1-based,
unit and partition numbers are 0-based and -1 is their wildcard marker.)
Also, after doing all the probing and choosing, print the final result as
"Booting from <disk spec>" where disk spec has all the wildcards resolved
and looks like familiar BSD slice-and-partition notation (disk0s3a, etc).
Alexander Motin [Sun, 22 Nov 2015 17:03:38 +0000 (17:03 +0000)]
Explicitly call SEND CHANGE REQUEST for pre-24xx chips in target mode.
While later firmware always registers for RSCN requests, older one does
it only in initiator mode. But in target mode there RSCN can be the only
way to detect gone intiator.
The r289895 revision did not accounted for the block containing the
requested page, when calculating the run of pages. Include the pages
before/after the requested page, that fit into the reqblock, into the
calculation.
Noted by: glebius
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Rick Macklem [Sat, 21 Nov 2015 23:55:46 +0000 (23:55 +0000)]
When the nfsd threads are terminated, the NFSv4 server state
(opens, locks, etc) is retained, which I believe is correct behaviour.
However, for NFSv4.1, the server also retained a reference to the xprt
(RPC transport socket structure) for the backchannel. This caused
svcpool_destroy() to not call SVC_DESTROY() for the xprt and allowed
a socket upcall to occur after the mutexes in the svcpool were destroyed,
causing a crash.
This patch fixes the code so that the backchannel xprt structure is
dereferenced just before svcpool_destroy() is called, so the code
does do an SVC_DESTROY() on the xprt, which shuts down the socket upcall.
Ian Lepore [Sat, 21 Nov 2015 23:30:47 +0000 (23:30 +0000)]
Update the imx5/imx6 cpu_reset() implementation based on a new understanding
of the SRS (software reset) bit in the watchdog control register. Despite
what the manual seems to imply, this bit DOES trigger an immediate reset, as
opposed to simply flagging the type of reset as software-triggered.
Alexander Motin [Sat, 21 Nov 2015 21:01:00 +0000 (21:01 +0000)]
Fix target mode with fabric for pre-24xx chips.
For those chips we are not receiving login events, adding initiators
based on ATIO requests. But there is no port ID in that structure, so
in fabric mode we have to explicitly fetch it from firmware to be able
to do normal scan after that.
Svatopluk Kraus [Sat, 21 Nov 2015 19:55:01 +0000 (19:55 +0000)]
Fix BUS_DMA_MIN_ALLOC_COMP flag logic. When bus_dmamap_t map is being
created for bus_dma_tag_t tag, bounce pages should be allocated
only if needed.
Before the fix, they were allocated always if BUS_DMA_COULD_BOUNCE flag
was set but BUS_DMA_MIN_ALLOC_COMP not. As bounce pages are never freed,
it could cause memory exhaustion when a lot of such tags together with
their maps were created.
Note that there could be more maps in one tag by current design.
However BUS_DMA_MIN_ALLOC_COMP flag is tag's flag. It's set after
bounce pages are allocated. Thus, they are allocated only for first
tag's map which needs them.
Warner Losh [Sat, 21 Nov 2015 16:37:11 +0000 (16:37 +0000)]
Document why we use -z nonexecstack in the Makefile since it
is so unusual. Turn off mis-match warnings for building uathload
because the firmware .o file is produced in a way that we
can't get to match exactly. This fixes the build on mips,
so stop excluding it from the build.
Andrew Turner [Sat, 21 Nov 2015 15:30:08 +0000 (15:30 +0000)]
Move more bus_space_* files to be built by files.arm. This leaves the
definition in a file.* file under sys/arm/arm in the few cases we need it
for non-fdt platforms.
Alexander Motin [Sat, 21 Nov 2015 10:22:01 +0000 (10:22 +0000)]
Add API to obtain primary enclosure name and ID for /dev/sesX devices.
sesX device number may change between reboots, so to properly identify
the instance we need more data. Name and ID reported here may mach ones
reported by SCSI device, but that is not really required by specs.
Justin Hibbits [Sat, 21 Nov 2015 06:03:46 +0000 (06:03 +0000)]
trunc_page() goes through unsigned long, which is too short.
sizeof(unsigned long) < sizeof(vm_paddr_t) on Book-E, which uses 36-bit
addressing. With this, a CCSR with a physical address above 4GB successfully
maps.
Marius Strobl [Sat, 21 Nov 2015 02:49:33 +0000 (02:49 +0000)]
Merge from r290547:
Since r289279 bufinit() uses mp_ncpus so adapt to what x86 does and
set this variable already in cpu_mp_setmaxid().
While at it, rename cpu_cpuid_prop() to cpu_portid_prop() as well as
the MD cpuid variable to portid to avoid confusion with the MI use
of "cpuid" and make some variable static/global in order to reduce
stack usage.
Marius Strobl [Sat, 21 Nov 2015 02:08:47 +0000 (02:08 +0000)]
Avoid a NULL pointer dereference in bounce_bus_dmamap_unload() when
the map has been created via bounce_bus_dmamem_alloc(). In that case
bus_dmamap_unload(9) typically isn't called during normal operation
but still should be during detach, cleanup from failed attach etc.
Adrian Chadd [Sat, 21 Nov 2015 00:22:47 +0000 (00:22 +0000)]
mips: teach the malta platform about extended memory.
Extended memory here is "physical memory above 256MB".
"memsize" in the environment only grows to 256MB; "ememsize" is the entire
memory range. Extended memory shows up at physical address 0x90000000.
This allows for malta64 VMs to be created with > 256MB RAM, all the way
up to 2GB RAM.
Tested:
* qemu-devel package; qemu-system-mips64 -m 2048 (and -m 256 to test the
no-ememsize case.)
Rick Macklem [Sat, 21 Nov 2015 00:15:41 +0000 (00:15 +0000)]
Revert r283330 since it broke directory caching in the client.
At this time I cannot see a way to fix directory caching when it
has partial blocks in the buffer cache, due to the fact that the
syscall's uio_offset won't stay the same as the lblkno * NFS_DIRBLKSIZ
offset.
While those tools are not needed anymore they are necessary to build FreeBSD 9
and 10. it does not hurt to keep those tools around until both 9 and 10 branch
become EOLed.
Modify colldef(1) to build after the change in the collation header, and ensure
it does produce the same collation definition it used to generate for 9 and 10