dteske [Sat, 21 Dec 2013 16:43:40 +0000 (16:43 +0000)]
Move the installer's keymap solution introduced by SVN r259468. The new
location of /etc/rc.local on the install media is more appropriate as it
knows serial vs. non-serial and can also do the change earlier (so that
even the initial Install dialog can benefit from the change).
imp [Sat, 21 Dec 2013 16:23:31 +0000 (16:23 +0000)]
Plumb the cn_grab and cn_ungrab routines down into the uart
clients. Mask RX interrupts while grabbed on the atmel serial
driver. This UART interrupts every character. When interrupts are
enabled at the mountroot> prompt, this means the ISR eats the
characters. Rather than try to create a cooperative buffering system
for the low level kernel console, instead just mask out the ISR. For
NS8250 and decsendents this isn't needed, since interrupts only happen
after 14 or more characters (depending on the fifo settings). Plumb
such that these are optional so there's no change in behavior for all
the other UART clients. ddb worked on this platform because all
interrupts were disabled while it was running, so this problem wasn't
noticed. The mountroot> issue has been around for a very very long
time.
dumbbell [Sat, 21 Dec 2013 15:40:36 +0000 (15:40 +0000)]
drm/ttm, drm/radeon: Replace EINTR/ERESTART by ERESTARTSYS...
... 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.
glebius [Sat, 21 Dec 2013 14:41:32 +0000 (14:41 +0000)]
In r248885 I have reduced size of fake uio resid that ng_ksocket(4) passes
to the soreceive(). This exposed a bug. When reading from a raw socket,
when our fake limit is depleted, we receive a truncated mbuf chain, with
m->m_pkthdr.len > m_length(m). The first problem is that MSG_TRUNC was not
handled. The second one is that we didn't reinit uio_resid in our endless
loop (neither flags), and if socket buffer contained several records, then
we quickly deplete our fake limit. The third bug, actually introduced in
r248885, is that MJUMPAGESIZE isn't enough to handle maximum packet that
ng_ksocket(4) can theoretically receive.
Changes:
- Reinit uio_resid and flags before every call to soreceive().
- Set maximum acceptable size of packet to IP_MAXPACKET. As for now the
module doesn't support INET6.
- Properly handle MSG_TRUNC return from soreceive().
PR: 184601
Submitted & tested by: Viktor Velichkin <avisom yandex.ru>
Sponsored by: Nginx, Inc.
emaste [Sat, 21 Dec 2013 13:58:55 +0000 (13:58 +0000)]
Support double-width characters in vt(9)
Normal and bold fonts each have a glyph map for single or left half-
glyphs, and right half glyphs. The flag TF_CJK_RIGHT in term_char_t
requests the right half-glyph.
Reviewed by: ed@
Sponsored by: The FreeBSD Foundation
dumbbell [Sat, 21 Dec 2013 12:55:42 +0000 (12:55 +0000)]
vga_pci: Improve boot display detection
The previous code was checking the "VGA Enable" bit on the video card's
parent PCI-to-PCI bridge only. This didn't work for the case where the
video card is attached to the root PCI bus (ie. the card has no parent
PCI-to-PCI bridge).
Now, the new code:
1. checks the "VGA Enable" bit on the parent bridge only if it's a
PCI-to-PCI bridge;
2. always checks the "I/O" and "Memory address space decoding" bits
on the video card itself.
However, vendor-specific bits are not used.
This fixes the use of many integrated Radeon cards: without this patch,
we fail to detect them as the boot display and, when radeonkms looks for
the Video BIOS, it skips the shadow copy made by the System BIOS. It
then fails to fully initialize the card, because the shadow copy is the
only way to read the Video BIOS in these situations. A workaround was to
force the boot display selection using the "hw.pci.default_vgapci_unit"
tunable.
A previous version of this patch added a new function doing the checks.
Now, the vga_pci_is_boot_display() function is used to perform the
checks (only until the boot display is found) and return if the given
device is the boot display or not.
Furthermore, vga_pci_attach() logs "Boot video device" if the card being
attached it the Chosen One:
vgapci0: <VGA-compatible display> [...]
vgapci0: Boot video device
jhibbits [Sat, 21 Dec 2013 00:07:56 +0000 (00:07 +0000)]
Add suspend/resume to DBDMA and ATA on PowerMacs.
This, and several subsequent commits, are suspend/resume for various PowerMac
drivers, which will include a change to the global suspend/resume code
eventually.
ed [Fri, 20 Dec 2013 21:31:50 +0000 (21:31 +0000)]
Extend libteken to support CJK fullwidth characters.
Introduce a new formatting bit (TF_CJK_RIGHT) that is set when putting a
cell that is the right part of a CJK fullwidth character. This will
allow drivers like vt(9) to support fullwidth characters properly.
emaste@ has a patch to extend vt(9)'s font handling to increase the
number of Unicode -> glyph maps from 2 ({normal,bold)} to 4
({normal,bold} x {left,right}). This will need to use this formatting
bit to determine whether to draw the left or right glyph.
gnn [Fri, 20 Dec 2013 20:22:10 +0000 (20:22 +0000)]
Add another Haswell model (0x45) to the set of supported chips.
Model 0x45 appears, for example, in late 2013 Mac Book Pro models
and is properly emulated by VMware.
glebius [Fri, 20 Dec 2013 19:45:51 +0000 (19:45 +0000)]
Move list of ttys handling from the allocating procedures, to the
device creation stage. A device creation can fail, and in that case
an entry already on the list will be freed.
pfg [Fri, 20 Dec 2013 18:09:34 +0000 (18:09 +0000)]
gcc: revert Rvalue references.
They are very useful but at this time I prefer not to figure out
some minor conflicts with the bigger Apple's blocks support
patch that is being worked on for current.
mav [Fri, 20 Dec 2013 17:39:07 +0000 (17:39 +0000)]
Remove several linear list traversals per request from RPC server code.
Do not insert active ports into pool->sp_active list if they are success-
fully assigned to some thread. This makes that list include only ports that
really require attention, and so traversal can be reduced to simple taking
the first one.
Remove idle thread from pool->sp_idlethreads list when assigning some
work (port of requests) to it. That again makes possible to replace list
traversals with simple taking the first element.
jhibbits [Fri, 20 Dec 2013 16:37:03 +0000 (16:37 +0000)]
Add suspend/resume handlers for ATI backlight ppc driver.
With this, also shut shut off the display (DPMS-style) and disable the clocking
when the backlight level is set to 0. This is taken from the radeonkms driver
(radeon_legacy_encoders.c) which doesn't yet support PowerPC, and won't for a
while, as it's missing full AGP support.
attilio [Fri, 20 Dec 2013 14:03:56 +0000 (14:03 +0000)]
o Remove assertions on ipa_version as sometimes the version detection
using cpuid can be quirky (this is the case of VMWare without the
vPMC support) but fail to probe hwpmc.
o Apply the fix for XEON family of processors as established by
315338-020 document (bug AJ85).
glebius [Fri, 20 Dec 2013 10:57:47 +0000 (10:57 +0000)]
ixgbe(4) takes packet counters from hardware in ixgbe_update_stats_counters(),
so we don't need to do a per packet increment, which trashes cache line.
adrian [Fri, 20 Dec 2013 07:41:03 +0000 (07:41 +0000)]
Disable the now unpredicably bogus check for whether we have
eneough queue space before queuing a bunch of IP fragments.
As the comment in the committed change says, in the post-if_transmit(),
post-SMP, post-preemption world, there's just too much overlapping
concurrent code paths and different approaches to driver transmit
queue management to have this code even remotely be effective.
The only specific place it could be useful is if ALTQ is enabled
but again it doesn't at all promise that all the fragments will be
transmitted anyway.
The main reason for committing this change is to disable a parallel
place where the drops counter is incremented. This is a side effect
of an upcoming change to ixgbe/cxgbe to handle the queue drops
counter slightly better.
neel [Fri, 20 Dec 2013 05:50:22 +0000 (05:50 +0000)]
Re-arrange bits in the amd64/pmap 'pm_flags' field.
The least significant 8 bits of 'pm_flags' are now used for the IPI vector
to use for nested page table TLB shootdown.
Previously we used IPI_AST to interrupt the host cpu which is functionally
correct but could lead to misleading interrupt counts for AST handler. The
AST handler was also doing a lot more than what is required for the nested
page table TLB shootdown (EOI and IRET).
ganbold [Fri, 20 Dec 2013 00:56:23 +0000 (00:56 +0000)]
Add identification and necessary type checks for Krait CPU cores. Krait CPU is used in
Qualcomm Snapdragon S4 and Snapdragon 400/600/800 SoCs and has architectural
similarities to ARM Cortex-A15. As for development boards IFC6400 series embedded
boards from Inforce Computing uses Snapdragon S4 Pro/APQ8064.
melifaro [Fri, 20 Dec 2013 00:17:26 +0000 (00:17 +0000)]
Use more fine-grained kvm(3) symbol lookup: routing code retrieves only
necessary symbols needed per subsystem. Main kvm(3) init is now delayed
as much as possbile. This finally fixes performance issues reported in
kern/167204.
Some non-working code (ng_socket.ko symbol addresses calculation) removed.
Some global variables eliminated.
mav [Thu, 19 Dec 2013 21:31:28 +0000 (21:31 +0000)]
Rework flow control for connection-oriented (TCP) RPC server.
When processing receive buffer, write the amount of data, expected
in present request record, into socket's so_rcv.sb_lowat to make stack
aware about our needs. When processing following upcalls, ignore them
until socket collect enough data to be read and processed in one turn.
This change reduces number of context switches and other operations
in RPC stack during large NFS writes (especially via non-Jumbo networks)
by order of magnitude.
After precessing current packet, take another look into the pending
buffer to find out whether the next packet had been already received.
If not, deactivate this port right there without making RPC code to
push this port to another thread just to find that there is nothing.
If the next packet is received partially, also deactivate the port, but
also update socket's so_rcv.sb_lowat to not be woken up prematurely.
This change additionally reduces number of context switches per NFS
request about in half.
se [Thu, 19 Dec 2013 09:01:46 +0000 (09:01 +0000)]
Fix overflow for timeout values of more than 68 years, which is the maximum
covered by sbintime (LONG_MAX seconds).
Some programs use timeout values in excess of 1000 years. The conversion
to sbintime caused wrap-around on overflow, which resulted in short or
negative timeout values. This caused long delays on sockets opened by
affected programs (e.g. OpenSSH).
Kernels compiled without -fno-strict-overflow were not affected, apparently
because the compiler tested the sign of the timeout value before performing
the multiplication that lead to overflow.
When the -fno-strict-overflow option was added to CFLAGS, this optimization
was disabled and the test was performed on the result of the multiplication.
Negative products were caught and resulted in EINVAL being returned, but
wrap-around to positive values just shortened the timeout value to the
residue of the result that could be represented by sbintime.
The fix is to cap the timeout values at the maximum that can be represented
by sbintime, which is 2^31 - 1 seconds or more than 68 years.
After this change, the kernel can be compiled with -fno-strict-overflow
with no ill effects.
dteske [Thu, 19 Dec 2013 06:41:30 +0000 (06:41 +0000)]
Fix a regression that occurred with SVN revisions 259474 and 259476 where
installation to 3-4+ (depending on vdev type) vdevs would result in odd
error messages where the zpool `create' command appeared to repeat itself
(an artifact of printf when you supply too many arguments -- caused by
neglecting to properly quote the multi-word expansion of $*vdevs when
creating the pool(s)). Example error below (taken from bsdinstall_log):
marcel [Thu, 19 Dec 2013 05:23:10 +0000 (05:23 +0000)]
Fix readdir for the root directory on a FAT32 file system. The root
directory is like any subdirectory and as such needs to use a real
cluster number. To this end, keep a DE structure for the root in
the DOS_FS structure and populate it accordingly.
While here:
o allow consecutive path separators by skipping them all.
o add missing $FreeBSD$ keyword to dosfs.h.
markj [Thu, 19 Dec 2013 03:48:36 +0000 (03:48 +0000)]
Invoke the kld_* event handlers from linker_load_file() and
linker_unload_file() rather than kern_kldload() and kern_kldunload(). This
ensures that the handlers are invoked for files that are loaded/unloaded
automatically as dependencies. Previously, they were only invoked for files
loaded by a user.
As a side effect, the kld_load and kld_unload handlers are now invoked with
the kernel linker lock exclusively held.
pjd [Thu, 19 Dec 2013 00:51:48 +0000 (00:51 +0000)]
If we cannot connect to casperd we don't enter sandbox, but if we can connect
to casperd, but we cannot access the service we need we exit with an error.
This should not happen and just indicates some configuration error which
should be fixed, so we force the user to do it by failing.
dteske [Wed, 18 Dec 2013 21:16:57 +0000 (21:16 +0000)]
In bsdinstall_log, it's rather confusing to see the following sequence:
DEBUG: Running installation step: hostname
rm: /tmp/bsdinstall_etc/fstab: No such file or directory
The two lines are unrelated, and the rm is spurious. Let's add `-f' to
that rm(1) so it doesn't confuse us when debugging an install.
dteske [Wed, 18 Dec 2013 21:13:50 +0000 (21:13 +0000)]
Fix "[: -eq: argument expected" error introduced by SVN r256343. The code
should not have used DIALOG_CANCEL because dialog.subr wasn't included to
define it. The effect of the error was that you could not cancel the
partition dialog. Discovered by checking bsdinstall_log for something else.
melifaro [Wed, 18 Dec 2013 20:17:05 +0000 (20:17 +0000)]
Add net.inet.ip.fw.dyn_keep_states sysctl which
re-links dynamic states to default rule instead of
flushing on rule deletion.
This can be useful while performing ruleset reload
(think about `atomic` reload via changing sets).
Currently it is turned off by default.
melifaro [Wed, 18 Dec 2013 18:25:27 +0000 (18:25 +0000)]
Switch netstat -rn to use standard API for retrieving list of routes
instead of peeking inside in-kernel radix via kget.
This permits us to change kernel structures without breaking userland.
Additionally, this change provide more reliable and faster output.
`Refs` and `Use` fields available in IPv4 by default (and via -W
for other families) were removed. `Refs` is radix-specific thing
which is not informative for users. `Use` field value is handy sometimes,
but a) current API does not support it and b) I'm not sure we will
support per-rte pcpu counters in near future.
Old method of retrieving data is still supported (either by defining
NewTree=0 or running netstat with -A). However, Refs/Use fields are
hidden.
marcel [Wed, 18 Dec 2013 17:03:43 +0000 (17:03 +0000)]
Fix an inappropriate free of a non-dynamic value. While here, make the
code more naive and robust:
1. When setting ev_value, also always set ev_flags appropriately
2. Always check ev_value and ev_flags before calling free.
Both the value and the EV_DYNAMIC property can come directly from the
consumers of the environment functionality, so it's good to be careful.
And since this code is typically not looked at for long periods of
time, it's good to have it be a little "dumb-looking".
Trigger case for the bug:
env_setenv("foo", 0, "1", NULL, NULL);
env_setenv("foo", 0, "2", NULL, NULL);
glebius [Wed, 18 Dec 2013 12:50:43 +0000 (12:50 +0000)]
- Rename tty_makedev() into tty_makedevf() and make it capable
to fail and return error.
- Use make_dev_p() in tty_makedevf() instead of make_dev_cred().
- Always pass MAKEDEV_CHECKNAME flag.
- Optionally pass MAKEDEV_REF flag.
- Provide macro for compatibility with old API.
This fixes races with simultaneous creation and desctruction of
ttys, and makes it possible to call tty_makedevf() from device
cloners.
A race in tty_watermarks() still exist, since the latter drops
lock for M_WAITOK allocation. This will be addressed in separate
commit.
yongari [Wed, 18 Dec 2013 07:23:42 +0000 (07:23 +0000)]
Failed m_devget(9) indicates lack of free mbuf cluster. Update
if_iqdrops counter for that case since the received frame is ok.
While here, simplify updating counter logic.
neel [Wed, 18 Dec 2013 06:24:21 +0000 (06:24 +0000)]
Use vmcs_read() and vmcs_write() in preference to vmread() and vmwrite()
respectively. The vmcs_xxx() functions provide inline error checking of
all accesses to the VMCS.
markj [Wed, 18 Dec 2013 01:41:52 +0000 (01:41 +0000)]
The fasttrap fork handler is responsible for removing tracepoints in the
child process that were inherited from its parent. However, this should
not be done in the case of a vfork, since the fork handler ends up removing
the tracepoints from the shared vm space, and userland DTrace probes in the
parent will no longer fire as a result.
Now the child of a vfork may trigger userland DTrace probes enabled in its
parent, so modify the fasttrap probe handler to handle this case and handle
the child process in the same way that it would handle the traced process.
In particular, if once traces function foo() in a process that vforks, and
the child calls foo(), fasttrap will treat this call as having come from the
parent. This is the behaviour of the upstream code.
While here, add #ifdef guards to some code that isn't present upstream.