Use a shadow buffer and never read from the frame buffer. Remove large slow
code for reading from the frame buffer.
Reading from the frame buffer is usually much slower than writing to
the frame buffer. Typically 10 to 100 times slower. It old modes,
it takes many more PIOs, and in newer modes with no PIOs writes are
often write-combined while reads remain uncached.
Reading from the frame buffer is not very common, so this change doesn't
give speedups of 10 to 100 times. My main test case is a floodfill()
function that reads about as many pixels as it writes. The speedups
are typically a factor of 2 to 4.
Duplicating writes to the shadow buffer is slower when no reads from the
frame buffer are done, but reads are often done for the pixels under the
mouse cursor, and doing these reads from the shadow buffer more than
compensates for the overhead of writing the shadow buffer in at least the
slower modes. Management of the mouse cursor also becomes simpler.
The shadow buffer doesn't take any extra memory, except twice as much
in old 4-plane modes. A buffer for holding a copy of the frame buffer
was allocated up front for use in the screen switching signal handler.
This wasn't changed when the handler was made async-signal safe. Use
the same buffer the shadow (but make it twice as large in the 4-plane
modes), and remove large special code for writing it as well as large
special code for reading ut. It used to have a rawer format in the
4-plane modes. Now it has a bitmap format which takes twice as much
memory but can be written almost as fast without special code.
VIDBUFs that are not the whole frame buffer were never supported, and the
change depends on this. Check for invalid VIDBUFs in some places and do
nothing. The removed code did something not so good.
adrian [Sun, 21 Apr 2019 02:36:01 +0000 (02:36 +0000)]
[ath] [ath_hal] [ath_hal_9300] Extend the start PCU receive to handle resetting ANI.
One of the fun issues with scanning has been how the existing
ANI values were programmed into the hardware when channels were
changed. If you're on a really crappy channel and ANI has made
you deaf then when you scan you continue to be deaf on all channels.
This code passes in a flag to startpcureceive which in AR5416 and later
is also used to enable ANI. This allows it to know if it's a normal
operation or a scan operation.
This fixes my situation at home where a temporary spot of a device
going deaf due to interference starts scanning and .. can't hear
anything until I restart.
Now, this isn't the full fix - ideally:
(a) all the ANI config and per-channel information would be migrated
to the shared HAL stuff and enabled for all of the NICs;
(b) when a station reassociates and some other error conditions
(like missed beacons, NF calibration failures, etc) a knob
to reset ANI parameters would likely help recovery.
psm(4): respect tap_disabled configuration with enabled Extended support
This fixes a bug where, even when hw.psm.tap_enabled=0, touchpad taps
were processed.
tap_enabled has three states: unconfigured, disabled, and enabled (-1, 0, 1).
To respect PR kern/139272, taps are ignored only when explicity disabled.
Make libvgl mostly work without superuser privilege in direct modes by
not doing any unnecessary PIO instructions or refusing to start when the
i/o privilege needed for these instructions cannot be acquired.
This turns off useless palette management in direct modes. Palette
management had no useful effect since the hardware palette is not used
in these modes.
This transiently acquires i/o privilege if possible as needed to give
VGLSetBorder() and VGLBlankDisplay() a chance of working. Neither has
much chance of working. I was going to drop support for them in direct
modes, but found that VGLBlankDisplay() still works with an old graphics
card on a not so old LCD monitor.
This has some good side effects: reduce glitches for managing the palette
for screen switches, and speed up and reduce async-signal-unsafeness in
mouse cursor drawing.
Replace hard tabs with four-character indentations, per PEP8.
This is being done to separate stylistic changes from the tests from functional
ones, as I accidentally introduced a bug to the tests when I used four-space
indentation locally.
Use symlinks for kernel modules rather than hardlinks
When aliasing a kernel module to a different name (ie if_igb for if_em),
it's better to use symlinks than hard links. kldxref will omit entries for
the links, ensuring that the loaded module has the correct name.
Reviewed by: imp
MFC after: 2 weeks
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D19979
There's a number of EFI_ZFS_BOOT #ifdefs that aren't needed, or can be
eliminated with some trivial #defines. Remove the EFI_ZFS_BOOT ifdefs
that aren't needed. Replace libzfs.h include which is not safe to
include without EFI_ZFS_BOOT with efizfs.h which is and now
conditionally included libzfs.h. Define efizfs_set_preferred away
and define efi_zfs_probe to NULL when ZFS is compiled out.
myjob will get loaded at the next second==0, but this echo job will not
run until cron restarts. These jobs are normally handled in
run_reboot_jobs(), which sets e->lastexit of INTERVAL jobs to the startup
time so they run 'n' seconds later.
Fix this by special-casing TargetTime > 0 in the database load. Preexisting
jobs will be handled at startup during run_reboot_jobs as normal, but if
we've reloaded a database during runtime we'll hit this case and set
e->lastexit to the current time when we process it. They will then run every
'n' seconds from that point, and a full restart of cron is no longer
required to make these jobs work.
Reported by: Juraj Lutter (otis_sk.freebsd.org)
Reviewed by: allanjude, bapt, bjk (earlier version), Juraj Lutter
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D19924
Since writes don't necessarily need to be on erase-block boundaries, we can
relax the block size and alignments down to sector size. If it needs to be
erased, opalflash_erase() will check proper alignment and size.
Add support for the ModeSetMasked attribute to the NFSv4.1 server.
I do not know of an extant NFSv4.1 client that currently does a Setattr
operation for the ModeSetMasked, but it has been discussed on the linux-nfs
mailing list.
This patch adds support for doing a Setattr of ModeSetMasked, so that it
will work for any future NFSv4.1 client that chooses to do so.
Tested via a hacked FreeBSD NFSv4.1 client.
Replace "vp" with NULL to make the code more readable.
At the time of this nfsv4_sattr() call, "vp == NULL", so this patch doesn't
change the semantics, but I think it makes the code more readable.
It also makes it consistent with the nfsv4_sattr() call a few lines above
this one. Found during code inspection.
Fix copying of overlapping bitmaps. The cases of copying within the
screen bitmap and within a single MEMBUF were broken when first source
line is before the first destination line and the sub-bitmaps overlap.
The fix just copies horizontal lines in reverse order when the first
source line is before the first destination line. This switches
directions unnecessarily in some cases, but the switch is about as
fast as doing a precise detection of overlaps. When the first lines
are the same, there can be undetected overlap in the horizontal
direction. The old code already handles this mostly accidentally by
using bcopy() for MEMBUFs and by copying through a temporary buffer
for the screen bitmap although the latter is sub-optimal in direct
modes.
cem [Fri, 19 Apr 2019 20:08:45 +0000 (20:08 +0000)]
kernel build: Disable unhelpful GCC warning (tripped after r346352)
-Wformat-zero-length does not highlight any particularly wrong code and it
is especially meaningless for device_printf(). Turn it off entirely to
remove a source of false positives.
cem [Fri, 19 Apr 2019 20:05:47 +0000 (20:05 +0000)]
libkern: Bring in arc4random_uniform(9) from libc
It is a useful arc4random wrapper in the kernel for much the same reasons as
in userspace. Move the source to libkern (because kernel build is
restricted to sys/, but userspace can include any file it likes) and build
kernel and libc versions from the same source file.
Copy the documentation from arc4random_uniform(3) to the section 9 page.
While here, add missing arc4random_buf(9) symlink.
Add wrapper functions to convert strings to EFI_DEVICE_PATH
In anticipation of new functionality, create routines to convert char *
and a CHAR16 * to a EFI_DEVICE_PATH
EFI_DEVICE_PATH *efi_name_to_devpath(const char *path);
EFI_DEVICE_PATH *efi_name_to_devpath16(CHAR16 *path);
void efi_devpath_free(EFI_DEVICE_PATH *dp);
The first two return an EFI_DEVICE_PATH for the passed in paths. The
third frees up the storage the first two return when the caller is
done with it.
Add UEFI definitions related to converting string to DEVICE_PATH
Add definitions from UEFI 2.7 Errata B standards doc for converting a
text string to a device path. Added clearly missing 'e' at the end of
Device to resolve mismatch in that document in
EFI_DEVICE_PATH_FROM_TEXT_PROTOCOL element names.
When an IPv6 packet is received for a raw socket which has the
IPPROTO_IPV6 level socket option IPV6_CHECKSUM enabled and the
checksum check fails, drop the message. Without this fix, an
ICMP6 message was sent indicating a parameter problem.
Thanks to bz@ for suggesting a way to simplify this fix.
When a checksum has to be computed for a received IPv6 packet because it
is requested by the application using the IPPROTO_IPV6 level socket option
IPV6_CHECKSUM on a raw socket, ensure that the packet contains enough
bytes to contain the checksum at the specified offset.
Avoid a buffer overwrite in rip6_output() when computing the checksum
as requested by the user via the IPPROTO_IPV6 level socket option
IPV6_CHECKSUM. The check if there are enough bytes in the packet to
store the checksum at the requested offset was wrong by 1.
Add a stat counter to track ipv6 atomic fragments. Atomic fragments can be
generated in response to invalid path MTU values, but are also a potential
attack vector and considered harmful (see RFC6946 and RFC8021).
While here add tracking of the atomic fragment counter to netstat and systat.
I believe previous ifdef caused NULL dereference in later zfs_log_create()
on attempt to create file inside directory belonging to ephemeral group
created on illumos, trying to write to log information about GID domain
of the newly created file, inheriting the ephemeral GID.
This patch reuses original illumos SGID code with exception that due to
lack of ID mapping code on FreeBSD ephemeral GID will turn into GID_NOBODY
by another ifdef inside zfs_fuid_map_id().
We cannot just assume that any name which ends with a letter is a group
That's not been true since we allowed renaming of network interfaces. It's also
not true for things like epair0a.
Try to retrieve the group members for the name to check, since we'll get ENOENT
if the group doesn't exist.
Fix the NFSv4.0 server so that it does not support NFSv4.1 attributes.
During inspection of a packet trace, I noticed that an NFSv4.0 mount
reported that it supported attributes that are only defined for NFSv4.1.
In practice, this bug appears to be benign, since NFSv4.0 clients will
not use attributes that were added for NFSv4.1.
However, this was not correct and this patch fixes the NFSv4.0 server
so that it only supports attributes defined for NFSv4.0.
It also adds a definition for NFSv4.1 attributes that can only be set,
although it is only defined as 0 for now.
This is anticipation of the addition of support for the NFSv4.1 mode+mask
attribute soon.
powerpc/powernv: Make erasing before writes optional
If the OPAL flash driver supports writing without erase, it adds a
'no-erase' property to the flash device node. Honor that property and don't
bother erasing if it exists.
The inp lock is not needed for testing the V6 flag as that flag is set
once when the inp is created and never changes. For non-TCP socket
options the lock is immediately dropped after checking that flag.
This just pushes the lock down to only be acquired for TCP socket
options.
This isn't a hot-path, more a cosmetic cleanup I noticed while reading
the code.
cem [Thu, 18 Apr 2019 20:48:54 +0000 (20:48 +0000)]
random(4): Restore availability tradeoff prior to r346250
As discussed in that commit message, it is a dangerous default. But the
safe default causes enough pain on a variety of platforms that for now,
restore the prior default.
Some of this is self-induced pain we should/could do better about; for
example, programmatic CI systems and VM managers should introduce entropy
from the host for individual VM instances. This is considered a future work
item.
On modern x86 and Power9 systems, this may be wholly unnecessary after
D19928 lands (even in the non-ideal case where early /boot/entropy is
unavailable), because they have fast hardware random sources available early
in boot. But D19928 is not yet landed and we have a host of architectures
which do not provide fast random sources.
This change adds several tunables and diagnostic sysctls, documented
thoroughly in UPDATING and sys/dev/random/random_infra.c.
Implement flag for telling cuse(3) clients if the peer is running in 32-bit
compat mode or not. This is useful when implementing compatibility ioctl(2)
handlers in userspace.
Add note that I'm the defacto maintainer of libegacy, our
compat shims for building bootstrap tools that use new features
on systems too old to have those new features. The world view
for this part of the build is unique to the whole build process
and continuty of design for new compat shims help this library
remain viable.
Correct handling of RMRR during early enumeration stages.
On some machines, DMAR contexts must be created before all devices
under the scope of the corresponding DMAR unit are enumerated.
Current code has two problems with that:
- scope lookup returns NULL device_t, which causes to skip creating a
context with RMRR, which is fatal for the affected device.
- calculation of the final pci dbsf address fails if any bridge in the
scope is not yet enumerated, because code relies on pcib_get_bus().
Make creation of contexts work either with device_t, or with DMAR PCI
scope paths. Scope provides enough information to infer context
address, and it is directly matched against DMAR tables scopes.
When calculating bus addresses for the scope or device, use direct
pci_cfgregread(PCIR_SECBUS_1) to get the secondary bus number, instead
of pcib_get_bus().
The issue was observed on HP Gen servers, where iLO PCI devices are
located behind south bridge switch. Turning on translation without
satisfying RMRR requests caused iLO to mostly hang, up to the level of
being unusable to control the server.
While there, remove hw.dmar.dmar_match_verbose tunable, and make the
normal logging under bootverbose useful and sufficient to diagnose
DRHD and RMRR parsing and matching.
FreeBSD does not set the pid field in the pfloghdr struct. This field is
populated on other platforms, document this to save people from trying
to use this field.
As an interim measure until a more permanent solution is implemented
workaround the following error:
/usr/src/contrib/elftoolchain/strings/strings.c:198:55: error: use of
undeclared identifier
'FA_OPEN' fa = fileargs_init(argc, argv, O_RDONLY, 0, &rights, FA_OPEN);
manu [Wed, 17 Apr 2019 21:45:19 +0000 (21:45 +0000)]
arm: allwinner: Fix audio for Allwinner H3/H5
Due to three conditions the codec driver for Allwinner A10/A20 and H3/H5 did not work properly here:
Wrong bit position for the analog audio reset
Hardware Reset of codec was not de-asserted correctly
Linux DTS file did not contain the address of the analog register the way as the driver was expecting it.
manu [Wed, 17 Apr 2019 20:09:01 +0000 (20:09 +0000)]
ofw_graph: Add functions for graph bindings
Those functions are helpers to work on graph bindings.
graphs are mostly use with video related devices.
See https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/devicetree/bindings/graph.txt?id=4436a3711e3249840e0679e92d3c951bcaf25515
sha1 is used by ether_gen_addr after r346324. Perhaps in an ideal world we
could detect that the kernel's been compiled without sha1_* bits included
and silently fallback to arc4random instead because these platforms/kernel
configs are far and few between. It's fairly lightweight, though, so just
include it for now.
Give devices that need a MAC a 16-bit allocation out of the FreeBSD
Foundation OUI range. Change the name ether_fakeaddr to ether_gen_addr now
that we're dealing real MAC addresses with a real OUI rather than random
locally-administered addresses.
If during DIOCRSETTFLAGS pfrio_buffer is NULL copyin() will fault, which we're
not allowed to do with a lock held.
We must count the number of entries in the table and release the lock during
copyin(). Only then can we re-acquire the lock. Note that this is safe, because
pfr_set_tflags() will check if the table and entries exist.
This was discovered by a local syzcaller instance.
ian [Wed, 17 Apr 2019 16:27:43 +0000 (16:27 +0000)]
Allow this test script to be run from within src/tools/boot dir, and create
the temporary image in $TMPDIR.
Allow the script to be run from the src/tools/boot directory by using make
-V SRCTOP to find the top of the tree, because this script is handy for
quick smoke-testing of loader changes, as well as being useful in CI testing.
Also, use a temp directory in $TMPDIR to assemble the boot image, and write
the boot log file to $TMPDIR. Arrange to have the temporary image clean
itself up, but leave the log file in $TMPDIR for post-mortem analysis of
failures when the script is run interactively.
Add fileargs_lstat function to cap_fileargs casper service to be able to
lstat files while in capability mode. It can only lstat files given in
fileargs_init.
Submitted by: Bora Özarslan <borako.ozarslan@gmail.com>
Reviewed by: oshogbo, cem (partial)
MFC after: 3 weeks
Relnotes: Yes
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D19548
ian [Wed, 17 Apr 2019 15:27:11 +0000 (15:27 +0000)]
Only set up the interrupts that will actually be used in arm generic_timer.
The code previously set up interrupt handlers for all the interrupt
resources available, including for timers that are not in use. That could
lead to interrupt storms. For example, if boot firmware enabled the virtual
timer but the kernel is using the physical timer, it could get flooded with
interrupts on the virtual timer which it cannot shut off. By only setting
up an interrupt handler for the hardware that will actually be used, any
interrupts from other timer units will remain masked in the interrupt
controller.
fdt: further consolidate DTB building and revise manpage
FDT_DTS_FILE was built separately with a rule in sys/conf/files and
recreated the rules we used in dtb.mk. Now that we have other infrastructure
to build a DTB along with the kernel, fold FDT_DTS_FILE into that since it
doesn't have any special requirements.
fdt(4) never got revised to mention the DTS/DTSO make options, so do that
now.
manu [Tue, 16 Apr 2019 20:08:19 +0000 (20:08 +0000)]
config: Only warn if duplicate option/device comes from the same file
This is useful for arm (possibly other arches too) where we want to have
a GENERIC kernel that only include files for the different SoC. Since
multiple SoCs/Board needs the same device we would need to do either :
Include the device in a generic file
Include the device in each file that really needs it
Option 1 works but if someone wants to create a specific kernel config
(which isn't uncommon for embedded system), he will need to add a lots
of nodevice to it.
Option 2 also works but produce a lots of warnings.
After the referenced commit, we did not set x87 and sse valid bits in
the xstate_bv bitmask for initial fpu state (stored in memory), when
using XSAVE.
The state is loaded into FPU register file to initialize the process
FPU state, and since both bits were clear, the default x87 and SSE
states were loaded. By chance, FreeBSD ABI SSE2 state is same as FPU
initial state, so the bug is not visible for 64bit processes. But on
i386, the precision control should be set to double (53bit mantissa),
instead of the default double extended (64bit mantissa). For 32bit
processes on amd64, kernel reloads control word with the right mask,
which only left native i386 and amd64 native but using x87 as
affected.
Fix it by setting minimal required xstate_bv mask.
Sponsored by: The FreeBSD Foundation
MFC after: 1 week
manu [Tue, 16 Apr 2019 19:38:16 +0000 (19:38 +0000)]
allwinner: clk: Garbage collect old clock implementation
The old clocks are disconneted from the build since r337344.
Remove all those pseudo drivers. The only one remaining is for gmac
(the ethernet controller) so move it to sys/arm/allwinner.
While here remove a83t support from gmacclk as it is unneeded since r326114.
cem [Tue, 16 Apr 2019 18:47:20 +0000 (18:47 +0000)]
stack_protector: Add tunable to bypass random cookies
This is a stopgap measure to unbreak installer/VM/embedded boot issues
introduced (or at least exposed by) in r346250.
Add the new tunable, "security.stack_protect.permit_nonrandom_cookies," in
order to continue boot with insecure non-random stack cookies if the random
device is unavailable.
For now, enable it by default. This is NOT safe. It will be disabled by
default in a future revision.
There is follow-on work planned to use fast random sources (e.g., RDRAND on
x86 and DARN on Power) to seed when the early entropy file cannot be
provided, for whatever reason. Please see D19928.
Some better hacks may be used to make the non-random __stack_chk_guard
slightly less predictable (from delphij@ and mjg@); those suggestions are
left for a future revision. I think it may also be plausible to move stack
guard initialization far later in the boot process; potentially it could be
moved all the way to just before userspace is started.
Reported by: many
Reviewed by: delphij, emaste, imp (all w/ caveat: this is a stopgap fix)
Security: yes
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D19927
cem [Tue, 16 Apr 2019 17:12:17 +0000 (17:12 +0000)]
random(4): Add is_random_seeded(9) KPI
The imagined use is for early boot consumers of random to be able to make
decisions based on whether random is available yet or not. One such
consumer seems to be __stack_chk_init(), which runs immediately after random
is initialized. A follow-up patch will attempt to address that.
Reported by: many
Reviewed by: delphij (except man page)
Approved by: secteam(delphij)
Differential Revision: https://reviews.freebsd.org/D19926
The cosqos field was added nearly 6 years ago in r254804, and it is
still unused by any in-tree consumers. I have a patchset that I'm
working on which aligns many network resources by NUMA domain,
including inps, inpcb lb group, tcp pacing, lagg output link
selection, backing pages for sendfile, and more. It reduces
cross-domain traffic by roughly 50% for a real web workload.
This patchset relies on being able to store the numa domain in the
mbuf, and grabbing the unused cosqos field for this purpose is the
first step in starting to usptream it.
Sigh, r346279 was also a test version with the reduced size doubled (so
it was actually double the full size in current kernels where the reduction
is null, so overran the mmapped() buffer).
Quick fix for slow clearing and context switches of large frame buffers
with old kernels, by breaking the support for large frame buffers in the
same way as for current kernels.
Large frame buffers may be too large to map into kva, and the kernel
(syscons) only uses the first screen page anyway, so r203535, r205557
and 248799 limit the buffer size in VESA modes to the first screen
page, apparently without noticing that this breaks applications by
using the same limit for user mappings as for kernel mappings. In
vgl, this makes the virtual screen the same as the physical screen.
However, this is almost a feature since clearing and switching large
(usually mostly unused) frame buffers takes too long. E.g., on a 16
year old low-end AGP card it takes about 12 seconds to clear the 128MB
frame buffer in old kernels that map it all and also map it with slow
attributes (e.g., uncacheable). Older PCI cards are even slower, but
usually have less memory. Newer PCIe cards are faster, but may have
many GB of memory. Also, vgl malloc()s a shadow buffer with the same
size as the frame buffer, so large frame buffers are even more wasteful
in applications than in the kernel.
Use the same limit in vgl as in newer kernels.
Virtual screens and panning still work in non-VESA modes that have
more than 1 page. The reduced buffer size in the kernel also breaks
mmap() of the last physical page in modes where the reduced size is
not a multiple of the physical page size. The same reduction in vgl
only reduces the virtual screen size.
projects / FreeBSD / FreeBSD.git / log