forced invalidation of the cache range regardless of the presence of
self-snoop feature. Some recent Intel GPUs in some modes are not
coherent, and dirty lines in CPU cache must be flushed before the
pages are transferred to GPU domain.
Reviewed by: alc (previous version)
Tested by: pho (amd64)
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
This feature is required by Mesa 9.2+. Without this, a GL application
crashes with the following message:
# glxinfo
name of display: :0.0
Gen6+ requires Kernel 3.6 or later.
Assertion failed: (ctx->Version > 0), function handle_first_current,
file ../../src/mesa/main/context.c, line 1498.
Abort (core dumped)
Now, Mesa 10.2.4 and 10.3-rc3 works fine:
# glxinfo
name of display: :0
display: :0 screen: 0
direct rendering: Yes
...
OpenGL renderer string: Mesa DRI Intel(R) 965GM
OpenGL version string: 2.1 Mesa 10.2.4
...
The code was imported from Linux 3.8.13.
Reviewed by: kib@
Tested by: kwm@, danfe@, Henry Hu,
Lundberg, Johannes <johannes@brilliantservice.co.jp>,
Johannes Dieterich <dieterich.joh@gmail.com>,
Lutz Bichler <lutz.bichler@gmail.com>,
MFC after: 3 days
Relnotes: yes
If DDB is active, we can't use a taskqueue thread to switch away from
the X window, because this thread can't run.
Reviewed by: ray@
Approved by: ray@
MFC after: 1 week
add opregion handling for drm2 - which exposes some ACPI video configuration
pieces that some Lenovo laptop models use to flesh out which video device
to speak to. This enables the brightness control in ACPI to work these models.
The CADL bits are also important - it's used to figure out which ACPI
events to hook the brightness buttons into. It doesn't yet seem to work
for me, but it does for the OP.
Tested:
* Lenovo X230 (mine)
* OP: ASUS UX51VZ
PR: 190186
Submitted by: Henry Hu <henry.hu.sh@gmail.com>
Reviewed by: dumbbell
vm_phys_fictitious_to_vm_page should not be called directly, even when
operating on a range that has been registered using
vm_phys_fictitious_reg_range. PHYS_TO_VM_PAGE should be used instead
because on arches that use VM_PHYSSEG_DENSE the page might come
directly from vm_page_array.
Reported by: nwhitehorn
Tested by: nwhitehorn, David Mackay <davidm.jx8p@gmail.com>
Sponsored by: Citrix Systems R&D
This change is a bit ugly, but so is the coupling between the i915
driver and syscons. It isn't worth developing a more elegant solution
only to support the legacy syscons console.
This includes:
o All directories named *ia64*
o All files named *ia64*
o All ia64-specific code guarded by __ia64__
o All ia64-specific makefile logic
o Mention of ia64 in comments and documentation
This excludes:
o Everything under contrib/
o Everything under crypto/
o sys/xen/interface
o sys/sys/elf_common.h
Discussed at: BSDcan
These changes prevent sysctl(8) from returning proper output,
such as:
1) no output from sysctl(8)
2) erroneously returning ENOMEM with tools like truss(1)
or uname(1)
truss: can not get etype: Cannot allocate memory
there is an environment variable which shall initialize the SYSCTL
during early boot. This works for all SYSCTL types both statically and
dynamically created ones, except for the SYSCTL NODE type and SYSCTLs
which belong to VNETs. A new flag, CTLFLAG_NOFETCH, has been added to
be used in the case a tunable sysctl has a custom initialisation
function allowing the sysctl to still be marked as a tunable. The
kernel SYSCTL API is mostly the same, with a few exceptions for some
special operations like iterating childrens of a static/extern SYSCTL
node. This operation should probably be made into a factored out
common macro, hence some device drivers use this. The reason for
changing the SYSCTL API was the need for a SYSCTL parent OID pointer
and not only the SYSCTL parent OID list pointer in order to quickly
generate the sysctl path. The motivation behind this patch is to avoid
parameter loading cludges inside the OFED driver subsystem. Instead of
adding special code to the OFED driver subsystem to post-load tunables
into dynamically created sysctls, we generalize this in the kernel.
Other changes:
- Corrected a possibly incorrect sysctl name from "hw.cbb.intr_mask"
to "hw.pcic.intr_mask".
- Removed redundant TUNABLE statements throughout the kernel.
- Some minor code rewrites in connection to removing not needed
TUNABLE statements.
- Added a missing SYSCTL_DECL().
- Wrapped two very long lines.
- Avoid malloc()/free() inside sysctl string handling, in case it is
called to initialize a sysctl from a tunable, hence malloc()/free() is
not ready when sysctls from the sysctl dataset are registered.
- Bumped FreeBSD version to indicate SYSCTL API change.
MFC after: 2 weeks
Sponsored by: Mellanox Technologies
the queue where to enqueue pages that are going to be unwired.
- Add stronger checks to the enqueue/dequeue for the pagequeues when
adding and removing pages to them.
Of course, for unmanaged pages the queue parameter of vm_page_unwire() will
be ignored, just as the active parameter today.
This makes adding new pagequeues quicker.
This change effectively modifies the KPI. __FreeBSD_version will be,
however, bumped just when the full cache of free pages will be
evicted.
Sponsored by: EMC / Isilon storage division
Reviewed by: alc
Tested by: pho
- Based on actual usage and on what Linux does, dummy_page.addr should
contain the physical bus address of the dummy page rather than its
virtual one. As a side-effect, correcting this bug fixes compilation
with PAE support enabled by getting rid of an inappropriate cast.
- Also based on actual usage of dummy_page.addr, theoretically Radeon
devices could do a maximum of 44-bit DMA. In reality, though, it is
more likely that they only support 32-bit DMA, at least that is what
radeon_gart_table_ram_alloc() sets up for, too. However, passing ~0
to drm_pci_alloc() as maxaddr parameter translates to 64-bit DMA on
amd64/64-bit machines. Thus, use BUS_SPACE_MAXSIZE_32BIT instead,
which the existing 32-bit DMA limits within the drm2 code spelled as
0xFFFFFFFF should also be changed to.
Reviewed by: dumbbell
MFC after: 1 week
Sponsored by: Bally Wulff Games & Entertainment GmbH
This allows to run 32bit applications on a 64bit host. This was tested
successfully with Wine (emulators/i386-wine-devel) and StarCraft II.
Submitted by: Jan Kokemüller <jan.kokemueller@gmail.com>
MFC after: 1 week
The priority goes from "error" to "debug".
Connectors are polled every 10 seconds. Reading EDID is part of this
polling. However, when an invalid EDID is returned, this error message
is logged. When using Newcons for instance, having a kernel message
every 10 seconds is getting annoying.
Now that it's a debug message, it'll be logged only if hw.dri.debug is
enabled. This fix console spamming for some users.
Tested by: Larry Rosenman <ler@lerctr.org>
... for msleep/cv_*wait() return values, where wait_event*() is used
on Linux. ERESTARTSYS is the return code expected by callers when the
operation was interrupted.
For instance, this is the case of radeon_cs_ioctl() (radeon_cs.c): if
an error occurs, and the code isn't ERESTARTSYS (eg. EINTR), it logs an
error.
Note that ERESTARTSYS is defined as ERESTART, but this keeps callers'
code close to Linux.
Submitted by: avg@ (previous version)
Previously the code would just iterate over the whole tree as if it were
just a list.
Without this change I would observe X server becoming more and more
jerky over time.
MFC after: 5 days
The code was unmodified compared to Linux and returned the amount of
received bytes from the i2c bus. This led to non-working i2c bus and
failure to eg. read monitor's EDID, if connected to DisplayPort.
MFC after: 3 days
Tested by: Mikaël Urankar <mikael.urankar@gmail.com>
This fixes radeon_agp_init() and gtt_size is now correct. However, this
is not enough to make Radeon AGP cards work: ttm_agp_backend.c isn't
implemented yet.
Submitted by: tijl@
The directory sys/dev/drm2/i915 is apperently contributed code.
Revert to the broken version of this file to make future imports easier.
Requested by: kib
shifts into the sign bit. Instead use (1U << 31) which gets the
expected result.
This fix is not ideal as it assumes a 32 bit int, but does fix the issue
for most cases.
A similar change was made in OpenBSD.
Discussed with: -arch, rdivacky
Reviewed by: cperciva
The code was easier to read without __DECONST and clang didn't report
any error. I thought the cast was enough...
MFC after: 3 days
X-MFC-With: r258549
drm_le_cmp() (qsort_r()'s callback) receives pointers to elements in the
array passed to qsort_r(), not the elements themselves.
Before this fix, the use of qsort_r() shuffled the array, not sorted it,
because the compare callback accessed random memory locations, not the
expected elements.
This bug triggered an infinite loop in KDE/xserver:
1. KDE has a kded module called "randrmonitor" which queries xserver
for current monitors at startup and then listens to RandR
notifications from xserver.
2. xserver handles the query from "randrmonitor" by polling the
video device using the "drm_mode_getconnector()" ioctl. This
ioctl returns a list of connectors and, for those with a
connected monitor, the available modes. Each modes list is sorted
by the kernel before returning. When xserver gets the connectors
list, it sorts the modes lists again.
In the case of this bug, when two modes are equal (in xserver's
compare function PoV), their order is kept stable (ie. the
kernel order is kept for those two modes). And because the list
was shuffled by the kernel, the order of two equal modes was
frequently changed in the final modes list in xserver.
3. xserver compares the returned connectors list with the list
obtained earlier. In particular, it compares the sorted
modes lists for each connector. If a property of a connector
changes (eg. modes), xserver sends a "RRNotify_OutputChange"
notification.
Because of the change of order between equal modes, xserver sent
a notification after each polling of the connectors.
4. "randrmonitor" receives a notification, triggered by its query. The
notification doesn't contain the new connectors list, therefore, it
asks for the new list using the same function: go back to step #2.
MFC after: 3 days
This fixes DPMS with KDE and radeonkms. Without this, the display would
freeze when the monitor is put into sleep state, and only resumes after
several dozens of minutes once the monitor is powered on again.
Tested by: Mathias Picker <Mathias.Picker@virtual-earth.de>
already. Also, according to the specs, GDRST register is not in the
power well, so the forcewake for reset status read is excessive for
this reason.
Use plain register read for waiting of the reset completion
notification, to avoid gt_lock recursion. Linux upstream did the
similar change, but their code was already restructured.
Reported by: ray
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
Calling those functions with the drmn device as argument causes a panic,
because it's not a direct child of pci$N. They must be called with the
vgapci device instead.
This fix is not enough to make suspend/resume work reliably.
Approved by: re (blanket)
vga_pci_(un)map_bios() takes a vgapci device as argument, not a drmn
one. This fixes a bug where the BIOS couldn't be mapped if the device
wasn't the boot display.
Approved by: re (kib; blanket for following drm2/radeon commits)
an address in the first 2GB of the process's address space. This flag should
have the same semantics as the same flag on Linux.
To facilitate this, add a new parameter to vm_map_find() that specifies an
optional maximum virtual address. While here, fix several callers of
vm_map_find() to use a VMFS_* constant for the findspace argument instead of
TRUE and FALSE.
Reviewed by: alc
Approved by: re (kib)
- Relax atomic_read() and atomic_set() macros. Linux does not require any
memory barrier. Also, these macros may be even reordered or optimized away
according to the API documentation:
https://www.kernel.org/doc/Documentation/atomic_ops.txt
