Robert Noland [Fri, 23 Oct 2009 18:44:53 +0000 (18:44 +0000)]
Correct some issues with zfs boot.
- Teach it to read gang blocks. (essentially untested)
If you see "ZFS: gang block detected!", please let
me know, so we can either remove the printf if it
works, or fix it if it doesn't.
- If multiple partitions exist on a disk, probe them all.
We also need to reset dsk->start to 0 to read the right
sector here.
- With GPT, we can have 128 partitions.
- If the bootfs property has ever been set on a pool
it seems that it never goes away. zpool won't allow
you to add to the pool with the bootfs property set.
However, if you clear the property back to default
we end up getting 0 for the object number and read
a bogus block pointer and fail to boot.
- Fix some error printfs. The printf in the loader is
only capable of c,s and u formats.
Jung-uk Kim [Fri, 23 Oct 2009 18:41:00 +0000 (18:41 +0000)]
- When we restore VESA state, try BIOS POST earlier. VESA restore state
function may not work properly if we don't. Turn off hardware cursor as
vesa_set_mode() does.
- Add VBE 3.0 specific fields in VESA mode structure and pack it. Note
the padding is 190 bytes although VBE 3.0 says 189 bytes. It must be wrong
because the size of structure becomes 255 bytes and the specification says
it must be 256 bytes in total. In fact, an example code in the spec. does
it right, though. While we are at it, fix some i386-isms.
- Remove state buffer size limitation. It is no longer necessary since
sys/compat/x86bios/x86bios.c r198251.
- Move int 0x10 vector test into vesa_bios_post() as we always do it anyway.
John Baldwin [Fri, 23 Oct 2009 15:14:54 +0000 (15:14 +0000)]
- Fix several off-by-one errors when using MAXCOMLEN. The p_comm[] and
td_name[] arrays are actually MAXCOMLEN + 1 in size and a few places that
created shadow copies of these arrays were just using MAXCOMLEN.
- Prefer using sizeof() of an array type to explicit constants for the
array length in a few places.
- Ensure that all of p_comm[] and td_name[] is always zero'd during
execve() to guard against any possible information leaks. Previously
trailing garbage in p_comm[] could be leaked to userland in ktrace
record headers via td_name[].
John Baldwin [Fri, 23 Oct 2009 15:09:51 +0000 (15:09 +0000)]
Don't bother copying the name of a kproc or kthread out into a temporary
array just to pass that array to printf(). kproc and kthread names are
NUL-terminated and can be printed using printf() directly.
Jilles Tjoelker [Fri, 23 Oct 2009 14:50:11 +0000 (14:50 +0000)]
wordexp(3): fix some bugs with signals and long outputs
* retry various system calls on EINTR
* retry the rest after a short read (common if there is more than about 1K
of output)
* block SIGCHLD like system(3) does (note that this does not and cannot
work fully in threaded programs, they will need to be careful with wait
functions)
Nathan Whitehorn [Fri, 23 Oct 2009 14:27:40 +0000 (14:27 +0000)]
Do not map the trap vectors into the kernel's address space. They are
only used in real mode and keeping them mapped only serves to make NULL
a valid address, which results in silent NULL pointer deferences.
Suggested by: Patrick Kerharo
Obtained from: projects/ppc64
Nathan Whitehorn [Fri, 23 Oct 2009 03:17:02 +0000 (03:17 +0000)]
Add SMP support on U3-based G5 systems. This does not yet work perfectly:
at least on my Xserve, getting the decrementer and timebase on APs to tick
requires setting up a clock chip over I2C, which is not yet done.
While here, correct the 64-bit tlbie function to set the CPU to 64-bit
mode correctly.
John Baldwin [Thu, 22 Oct 2009 14:53:44 +0000 (14:53 +0000)]
Set the devclass_t pointer specified in the DRIVER_MODULE() macro
sooner so it is always valid when a driver's identify routine is
called. Previously, new-bus would attempt to create the devclass for
a newly loaded driver in two separate places, once in
devclass_add_driver(), and again after devclass_add_driver() returned
in driver_module_handler(). Only the second lookup attempted to set a
device class' parent and set the devclass_t pointer specified in the
DRIVER_MODULE() macro. However, by the time it was executed, the
driver was already added to existing instances of the parent driver at
which point in time the new driver's identify routine would have been
invoked. The fix is to merge the two attempts and only create the
devclass once in devclass_add_driver() including setting the
devclass_t pointer passed to DRIVER_MODULE() before the driver is
added to any existing bus devices.
Philip Paeps [Wed, 21 Oct 2009 23:50:35 +0000 (23:50 +0000)]
Make dhclient use bootpc (68) as the source port for unicast DHCPREQUEST
packets instead of allowing the protocol stack to pick a random source port.
This fixes the behaviour where dhclient would never transition from RENEWING
to BOUND without going through REBINDING in networks which are paranoid about
DHCP spoofing, such as most mainstream cable-broadband ISP networks.
Reviewed by: brooks
Obtained from: OpenBSD (partly - I'm not convinced their solution can work)
MFC after: 1 week (pending re approval)
John Baldwin [Wed, 21 Oct 2009 19:26:12 +0000 (19:26 +0000)]
Change gcc to assume a default machine architecture of 486 instead of 386
on "i386". Doing it in the compiler is deemed to be less fragile then
attempting to provide a default -march setting via bsd.cpu.mk. FreeBSD
itself has not supported plain 386 CPUs since 5.x.
Suggested by: kan
Requested by: rdivacky
MFC after: 1 month
Marcel Moolenaar [Wed, 21 Oct 2009 18:38:02 +0000 (18:38 +0000)]
o Introduce vm_sync_icache() for making the I-cache coherent with
the memory or D-cache, depending on the semantics of the platform.
vm_sync_icache() is basically a wrapper around pmap_sync_icache(),
that translates the vm_map_t argumument to pmap_t.
o Introduce pmap_sync_icache() to all PMAP implementation. For powerpc
it replaces the pmap_page_executable() function, added to solve
the I-cache problem in uiomove_fromphys().
o In proc_rwmem() call vm_sync_icache() when writing to a page that
has execute permissions. This assures that when breakpoints are
written, the I-cache will be coherent and the process will actually
hit the breakpoint.
o This also fixes the Book-E PMAP implementation that was missing
necessary locking while trying to deal with the I-cache coherency
in pmap_enter() (read: mmu_booke_enter_locked).
The key property of this change is that the I-cache is made coherent
*after* writes have been done. Doing it in the PMAP layer when adding
or changing a mapping means that the I-cache is made coherent *before*
any writes happen. The difference is key when the I-cache prefetches.
Ed Schouten [Wed, 21 Oct 2009 18:31:54 +0000 (18:31 +0000)]
Make input parsing in Farhenheit actually work.
Don't clobber *p with '\0' when testing whether it has the value of 'F'.
Just use the semantics of strtof() properly. If it returns p, we know
that it parsed the string until it reached 'C' or 'F'.
The code has not changed since it has been imported (r161951, Sep 3,
2006).
Marcel Moolenaar [Wed, 21 Oct 2009 18:09:48 +0000 (18:09 +0000)]
o Align function on a 32-byte boundary so that the core's front-end
can deliver 2 bundles per cycle to the back-end.
o Mark syscall stubs with a special unwind ABI tag so that unwind
libraries know how to unwind.
Alexander Motin [Wed, 21 Oct 2009 15:27:48 +0000 (15:27 +0000)]
MFp4:
Separate CAM_DEV_IDENTIFY_DATA_VALID flag from CAM_DEV_INQUIRY_DATA_VALID.
Add workaround for very old devices without support for mode setting.
Add some PATA bus scanning support.
Remove some SCSIsms.
Alexander Motin [Wed, 21 Oct 2009 14:20:55 +0000 (14:20 +0000)]
MFp4:
Add support for PIO-only devices.
Fix maxio values and 256 sectors transactions for 28bits commands.
Implement periodic ordered commands insertion, sames as da driver does.
Remove some SCSIsms.
Stanislav Sedov [Wed, 21 Oct 2009 12:47:09 +0000 (12:47 +0000)]
- Introduce new option BCE_JUMBO_HDRSPLIT that allows user to enable header
splitting in bce(4) instead of (ab)using ZERO_COPY_SOCKETS that was not
propagated into if_bce.c anyway. It is disabled by default.
Alexander Motin [Wed, 21 Oct 2009 12:42:25 +0000 (12:42 +0000)]
MFp4:
On error, freeze device queue, to allow periph driver to do proper recovery.
Freeze SIM queue only in some cases, when it is needed to protect SIM.
Implement better command timeout detection logic for non-queued commands.
This fixes false positives when command with short timeout waiting for the
long one. For example, when hald tastes CD during burning process.
Stanislav Sedov [Wed, 21 Oct 2009 11:50:18 +0000 (11:50 +0000)]
- On entrance to the rx_eof sync RX rings maps with POSTWRITE flag
instead of POSTREAD: the hardware do not touch this memory (CPU
updates it). It is already synchronized as PREWRITE after the
processing is done.
- Synchronize RX return ring memory in rx_eof. This is needed
as the deviced updates this memory when receives packets.
- Decouple the synchronization of BGE status block in the interrupt
service routine: perfrom PREREAD synchronization only all accesses
to this block are finished. This seems to be more natural.
Andrew Thompson [Tue, 20 Oct 2009 21:29:46 +0000 (21:29 +0000)]
Change from CAM_TID_INVALID to CAM_SEL_TIMEOUT error code when the usb device
has been yanked, this works around a cam recounting bug when
CAM_DEV_UNCONFIGURED is set late in the detach. In certain conditions the
reference to the XPT device would not be released which would cause the usb
explore thread to sleep forever on "simfree", preventing any new usb devices to
be found/ejected on the bus.
This is intended to be a quick workaround to the problem without touching CAM
so it can be merged to 8.0.
Qing Li [Tue, 20 Oct 2009 21:27:03 +0000 (21:27 +0000)]
The flow-table function flowtable_route_flush() may be called
during system initialization time. Since the flow-table is
designed to maintain per CPU flow cache, the existing code
did not check whether "smp_started" is true before calling
sched_bind() and sched_unbind(), which triggers a page fault.
Andrew Gallatin [Tue, 20 Oct 2009 18:58:28 +0000 (18:58 +0000)]
Make mxge do a better job recovering from NIC h/w faults
by checking PCI config space when the NIC is not
transmitting. Previously, a h/w fault would not have been
detected if the NIC was down, or handling an RX only
workload.
Qing Li [Tue, 20 Oct 2009 17:55:42 +0000 (17:55 +0000)]
In the ARP callout timer expiration function, the current time_second
is compared against the entry expiration time value (that was set based
on time_second) to check if the current time is larger than the set
expiration time. Due to the +/- timer granularity value, the comparison
returns false, causing the alternative code to be executed. The
alternative code path freed the memory without removing that entry
from the table list, causing a use-after-free bug.
Ruslan Ermilov [Tue, 20 Oct 2009 16:36:51 +0000 (16:36 +0000)]
Random number generator initialization cleanup:
- Introduce new SI_SUB_RANDOM point in boot sequence to make it
clear from where one may start using random(9). It should be as
early as possible, so place it just after SI_SUB_CPU where we
have some randomness on most platforms via get_cyclecount().
- Move stack protector initialization to be after SI_SUB_RANDOM
as before this point we have no randomness at all. This fixes
stack protector to actually protect stack with some random guard
value instead of a well-known one.
Note that this patch doesn't try to address arc4random(9) issues.
With current code, it will be implicitly seeded by stack protector
and hence will get the same entropy as random(9). It will be
securely reseeded once /dev/random is feeded by some entropy from
userland.
Submitted by: Maxim Dounin <mdounin@mdounin.ru>
MFC after: 3 days
Jaakko Heinonen [Tue, 20 Oct 2009 15:06:18 +0000 (15:06 +0000)]
Unloading of the nfscl module is unsupported because newnfslock doesn't
support unloading. It's not trivial to implement newnfslock unloading so
for now just admit that unloading is unsupported and refuse to attempt
unload in all nfscl module event handlers.
Jaakko Heinonen [Tue, 20 Oct 2009 15:01:46 +0000 (15:01 +0000)]
Fix ordering of nfscl_modevent() and ncl_uninit(). nfscl_modevent() must
be called after ncl_uninit() when unloading the nfscl module because
ncl_uninit() uses ncl_iod_mutex which is destroyed in nfscl_modevent().
Edwin Groothuis [Tue, 20 Oct 2009 06:54:31 +0000 (06:54 +0000)]
Instead of having to know which timezone was picked last time, you
now can run "tzsetup -r" which will reinstall the last choice. This
data is recorded in /var/db/zoneinfo.
Alexander Kabaev [Tue, 20 Oct 2009 02:35:12 +0000 (02:35 +0000)]
Use callout_init_mtx on FreeBSD versions recent enough. This closes
the race where interrupt thread can complete the request for which
timeout has fired and while mpt_timeout has blocked on mpt_lock.
Do a best effort to keep 4.x ang Giant-locked configurartions
compiling still.
Jung-uk Kim [Mon, 19 Oct 2009 20:58:10 +0000 (20:58 +0000)]
Rewrite x86bios and update its dependent drivers.
- Do not map entire real mode memory (1MB). Instead, we map IVT/BDA and
ROM area separately. Most notably, ROM area is mapped as device memory
(uncacheable) as it should be. User memory is dynamically allocated and
free'ed with contigmalloc(9) and contigfree(9). Remove now redundant and
potentially dangerous x86bios_alloc.c. If this emulator ever grows to
support non-PC hardware, we may implement it with rman(9) later.
- Move all host-specific initializations from x86emu_util.c to x86bios.c and
remove now unnecessary x86emu_util.c. Currently, non-PC hardware is not
supported. We may use bus_space(9) later when the KPI is fixed.
- Replace all bzero() calls for emulated registers with more obviously named
x86bios_init_regs(). This function also initializes DS and SS properly.
- Add x86bios_get_intr(). This function checks if the interrupt vector is
available for the platform. It is not necessary for PC-compatible hardware
but it may be needed later. ;-)
- Do not try turning off monitor if DPMS does not support the state.
- Allocate stable memory for VESA OEM strings instead of just holding
pointers to them. They may or may not be accessible always. Fix a memory
leak of video mode table while I am here.
- Add (experimental) BIOS POST call for vesa(4). This function calls VGA
BIOS POST code from the current VGA option ROM. Some video controllers
cannot save and restore the state properly even if it is claimed to be
supported. Usually the symptom is blank display after resuming from suspend
state. If the video mode does not match the previous mode after restoring,
we try BIOS POST and force the known good initial state. Some magic was
taken from NetBSD (and it was taken from vbetool, I believe.)
- Add a loader tunable for vgapci(4) to give a hint to dpms(4) and vesa(4)
to identify who owns the VESA BIOS. This is very useful for multi-display
adapter setup. By default, the POST video controller is automatically
probed and the tunable "hw.pci.default_vgapci_unit" is set to corresponding
vgapci unit number. You may override it from loader but it is very unlikely
to be necessary. Unfortunately only AGP/PCI/PCI-E controllers can be
matched because ISA controller does not have necessary device IDs.
- Fix a long standing bug in state save/restore function. The state buffer
pointer should be ES:BX, not ES:DI according to VBE 3.0. If it ever worked,
that's because BX was always zero. :-)
- Clean up register initializations more clearer per VBE 3.0.
- Fix a lot of style issues with vesa(4).
Powercrypt and NetSec seem to be defunct (webpages point to link farms
and a google search yields no alternative). Remove the links but
keep the entries around for reference.
PR: 139756
Submitted by: Patrick Oonk <patrick@pine.nl>
MFC after: 3 days