bdrewery [Thu, 7 Jan 2016 00:19:03 +0000 (00:19 +0000)]
Move the MAKEOBJDIRPREFIX value guard to sys.mk and expand to MAKEOBJDIR.
This will ensure that the variable was not set as a make override, in
make.conf, src.conf or src-env.conf. It allows setting the value in
src-env.conf when using WITH_AUTO_OBJ since that case properly handles
changing .OBJDIR (except if MAKEOBJDIRPREFIX does not yet exist which is
being discussed to be changed).
This change allows setting a default MAKEOBJDIRPREFIX via local.sys.env.mk.
emaste [Thu, 7 Jan 2016 00:15:02 +0000 (00:15 +0000)]
Switch GNU ld to be installed as ld.bfd and linked as ld
We intend to replace GNU ld with LLVM's lld, and on the path to there
we'll experiment with having lld installed or linked as /usr/bin/ld.
Thus, make ld.bfd the primary install target for GNU ld, to later
facilitate making the ld link optional.
Reviewed by: davide, dim
Differential Revision: https://reviews.freebsd.org/D4790
glebius [Thu, 7 Jan 2016 00:14:42 +0000 (00:14 +0000)]
Historically we have two fields in tcpcb to describe sender MSS: t_maxopd,
and t_maxseg. This dualism emerged with T/TCP, but was not properly cleaned
up after T/TCP removal. After all permutations over the years the result is
that t_maxopd stores a minimum of peer offered MSS and MTU reduced by minimum
protocol header. And t_maxseg stores (t_maxopd - TCPOLEN_TSTAMP_APPA) if
timestamps are in action, or is equal to t_maxopd otherwise. That's a very
rough estimate of MSS reduced by options length. Throughout the code it
was used in places, where preciseness was not important, like cwnd or
ssthresh calculations.
With this change:
- t_maxopd goes away.
- t_maxseg now stores MSS not adjusted by options.
- new function tcp_maxseg() is provided, that calculates MSS reduced by
options length. The functions gives a better estimate, since it takes
into account SACK state as well.
emaste [Wed, 6 Jan 2016 17:33:32 +0000 (17:33 +0000)]
Add fls() to libstand
Although we don't use it in tree yet libstand is installed as user-
facing /usr/liblibstand.a, and some work in progress makes use of it.
Instead of conflicting with ongoing libstand Makefile deduplication,
just add it now.
imp [Wed, 6 Jan 2016 17:13:40 +0000 (17:13 +0000)]
Try a little harder to remove firstboot and firstboot-reboot files in
case they accidentally get created as directories or with flags that
prevent their removal. While I wouldn't normally go the extra mile
here and let the normal unix rules prevail, the effects of failure are
large enough that extra care is warranted.
gjb [Wed, 6 Jan 2016 05:23:25 +0000 (05:23 +0000)]
Add a new target to touch the ${.OBJDIR}/release file, which
indicates the 'release' target has run (in order to prevent
subsequent invocations that may clobber original build output).
As is, the 'release' target is a dummy target that does nothing
more than depend on subsequent targets. Unless 'make obj' is
invoked prior to 'make release', .OBJDIR and .CURDIR will always
be '/usr/src/release' (or wherever /usr/src is located).
When 'make release' invokes 'make real-release' (and subsequent
targets), .OBJDIR is not updated, which still leads to src/ tree
pollution.
While arguably a hack, 'make release' will invoke the original
dummy targets as originally intended, but instead of touching an
empty file (or returing @true), will call a 'release-done' target
that will trigger the behavior that was intended to prevent
a subsequent invocation.
Discussed with: hrs
MFC after: 3 days
X-MFC-With: r293173
Sponsored by: The FreeBSD Foundation
asomers [Wed, 6 Jan 2016 00:00:11 +0000 (00:00 +0000)]
"source routing" in rpcbind
Fix a bug in rpcbind for multihomed hosts. If the server had interfaces on
two separate subnets, and a client on the first subnet contacted rpcbind at
the address on the second subnet, rpcbind would advertise addresses on the
first subnet. This is a bug, because it should prefer to advertise the
address where it was contacted. The requested service might be firewalled
off from the address on the first subnet, for example.
usr.sbin/rpcbind/check_bound.c
If the address on which a request was received is known, pass that
to addrmerge as the clnt_uaddr parameter. That is what addrmerge's
comment indicates the parameter is supposed to mean. The previous
behavior is that clnt_uaddr would contain the address from which the
client sent the request.
usr.sbin/rpcbind/util.c
Modify addrmerge to prefer to use an IP that is equal to clnt_uaddr,
if one is found. Refactor the relevant portion of the function for
clarity, and to reduce the number of ifdefs.
etc/mtree/BSD.tests.dist
usr.sbin/rpcbind/tests/Makefile
usr.sbin/rpcbind/tests/addrmerge_test.c
Add unit tests for usr.sbin/rpcbind/util.c:addrmerge.
usr.sbin/rpcbind/check_bound.c
usr.sbin/rpcbind/rpcbind.h
usr.sbin/rpcbind/util.c
Constify some function arguments
gjb [Tue, 5 Jan 2016 21:05:17 +0000 (21:05 +0000)]
Merge ^/projects/release-install-debug:
- Rework MANIFEST generation and parsing via bsdinstall(8).
- Allow selecting debugging distribution sets during install.
- Rework bsdinstall(8) to fetch remote debug distribution sets
when they are not available on the local install medium.
- Allow selecting additional non-GENERIC kernels during install.
At present, GENERIC is still required, and installed by default.
Tested with: head@r293203
Sponsored by: The FreeBSD Foundation
jilles [Tue, 5 Jan 2016 16:21:20 +0000 (16:21 +0000)]
Add sbin and /usr/local directories to _PATH_DEFPATH.
Set _PATH_DEFPATH to
/sbin:/bin:/usr/sbin:/usr/bin:/usr/local/sbin:/usr/local/bin. This is the
path in the default class in the default /etc/login.conf,
excluding ~/bin which would not be expanded properly in a string
constant.
For normal logins, _PATH_DEFPATH is overridden by /etc/login.conf,
~/.login_conf or shell startup files. _PATH_DEFPATH is still used as a
default by execlp(), execvp(), posix_spawnp() and sh if PATH is not set, and
by cron. Especially the latter is a common trap (most recently in PR
204813).
kib [Tue, 5 Jan 2016 14:48:40 +0000 (14:48 +0000)]
Two fixes for excessive iterations after r292326.
Advance the logical block number to the lblkno of the found block plus
one, instead of incrementing the block number which was used for
lookup. This change skips sparcely populated buffer ranges, similar
to r292325, instead of doing useless lookups.
Do not restart the bnoreuselist() from the start of the range if
buffer lock cannot be obtained without sleep. Only retry lookup and
lock for the same queue and same logical block number.
Reported by: benno
Tested by: pho
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
gjb [Tue, 5 Jan 2016 03:20:45 +0000 (03:20 +0000)]
Prevent memstick installation medium from attempting to mount
the root filesystem read-write. This causes problems booting
the memstick installation medium from write-protected USB flash
drives.
Submitted by: A.J. Kehoe IV [1], Oliver Jones [2]
PR: 187161 [1], 205886 [2]
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
iw_cxgbe: Shut down the socket but do not close the fd in case of error.
The fd is closed later in this case. This fixes a "SS_NOFDREF on enter"
panic.
Submitted by: Krishnamraju Eraparaju @ Chelsio
Reviewed by: Steve Wise @ Open Grid Computing
avos [Mon, 4 Jan 2016 21:11:27 +0000 (21:11 +0000)]
iwn: reduce code duplication in iwn_read_firmware()
- Separate 'firmware_put(sc->fw_fp, FIRMWARE_UNLOAD); sc->fw_fp = NULL;'
into iwn_unload_firmware().
- Move error handling to the end of iwn_read_firmware().
No functional changes.
Approved by: adrian (mentor)
Differential Revision: https://reviews.freebsd.org/D4768
melifaro [Mon, 4 Jan 2016 15:03:20 +0000 (15:03 +0000)]
Add rib_lookup_info() to provide API for retrieving individual route
entries data in unified format.
There are control plane functions that require information other than
just next-hop data (e.g. individual rtentry fields like flags or
prefix/mask). Given that the goal is to avoid rte reference/refcounting,
re-use rt_addrinfo structure to store most rte fields. If caller wants
to retrieve key/mask or gateway (which are sockaddrs and are allocated
separately), it needs to provide sufficient-sized sockaddrs structures
w/ ther pointers saved in passed rt_addrinfo.
Convert:
* lltable new records checks (in_lltable_rtcheck(),
nd6_is_new_addr_neighbor().
* rtsock pre-add/change route check.
* IPv6 NS ND-proxy check (RADIX_MPATH code was eliminated because
1) we don't support RTF_ANNOUNCE ND-proxy for networks and there should
not be multiple host routes for such hosts 2) if we have multiple
routes we should inspect them (which is not done). 3) the entire idea
of abusing KRT as storage for ND proxy seems odd. Userland programs
should be used for that purpose).
ngie [Mon, 4 Jan 2016 03:02:44 +0000 (03:02 +0000)]
Rename nitems and string variables to avoid collisions
Rename the `nitems` variable to `num_items` to avoid collisions with the
macro in sys/param.h for counting elements in an array
Similarly, rename `string` to `string_arr` to avoid future collisions with
potential keywords, as well as make it clear that `string_arr` isn't a char*
value, but instead a char** value.
jhibbits [Mon, 4 Jan 2016 02:20:14 +0000 (02:20 +0000)]
Make arguments for booke_init() u_long, to match register width.
On powerpc64, pointers are 64 bits, so casting from uint32_t changes the integer
width.
The alternative was to use register_t, but I didn't see register_t used as
argument type for any other functions, though didn't look too closely. u_long
was an acceptable alternative. On 64-bit it's 64 bits, on 32-bit it's 32 bits.
jhibbits [Mon, 4 Jan 2016 01:33:07 +0000 (01:33 +0000)]
Set the cacheline size before calling powerpc_init()
powerpc_init() initializes the mmu. Since this may clear pages via
pmap_zero_page(), set the cacheline size before calling into it, so
pmap_zero_page() has the right cacheline size. This isn't completely
necessary now, but will be when 64-bit book-e is completed.
imp [Sun, 3 Jan 2016 19:18:48 +0000 (19:18 +0000)]
Use /bin/rm to remove /firstboot*. Otherwise rm -i alias is picked
up and can cause issues on boot with the prompts. Fix the read-only
root case with horrible kludge of mounting rw removing the files, then
mounting ro. But since that's no more horrible than the kludge of
using marker files in /. With this change, NanoBSD configs can safely
use /firstboot + growfs to produce minimal images that grow to the
size of the card.
ngie [Sun, 3 Jan 2016 18:09:46 +0000 (18:09 +0000)]
Fix ixl(4) compilation with PCI_IOV pre-r266974
stable/10 doesn't have the if_getdrvflags(9) KPI. Reference the field in the
structure directly if the __FreeBSD_version is < 1100022, so the driver can
be built with PCI_IOV support on stable/10, without backporting all of
r266974 (which requires additional changes due to projects/ifnet, etc)
adrian [Sun, 3 Jan 2016 17:58:11 +0000 (17:58 +0000)]
[ath] remove the inline version of the register access macros.
These are going to be much more efficient on low end embedded systems
but unfortunately they make it .. less convenient to implement correct
bus barriers and debugging. They also didn't implement the register
serialisation workaround required for Owl (AR5416.)
So, just remove them for now. Later on I'll just inline the routines
from ah_osdep.c.
ian [Sun, 3 Jan 2016 14:46:19 +0000 (14:46 +0000)]
Eliminate code for walking through the early static env data. This code
is called from a device attach routine, and thus cannot be called before
the cutover from static to dynamic kernel env.
melifaro [Sun, 3 Jan 2016 10:43:23 +0000 (10:43 +0000)]
Remove 'struct route_int6' argument from in6_selectsrc() and
in6_selectif().
The main task of in6_selectsrc() is to return IPv6 SAS (along with
output interface used for scope checks). No data-path code uses
route argument for caching. The only users are icmp6 (reflect code),
ND6 ns/na generation code. All this fucntions are control-plane, so
there is no reason to try to 'optimize' something by passing cached
route into to ip6_output(). Given that, simplify code by eliminating
in6_selectsrc() 'struct route_in6' argument. Since in6_selectif() is
used only by in6_selectsrc(), eliminate its 'struct route_in6' argument,
too. While here, reshape rte-related code inside in6_selectif() to
free lookup result immediately after saving all the needed fields.
avos [Sun, 3 Jan 2016 10:06:10 +0000 (10:06 +0000)]
iwm: reorganize if_iwmvar.h
- Change order of data in if_iwmvar.h
(like it is in other drivers: defines, data structures,
vap/node structures, softc struct and locks); use indentation.
- Fix IWM_LOCK(_sc) / IWM_UNLOCK(_sc) macro.
- Add IWM_LOCK_INIT / DESTROY(sc) + fix mtx_init() usage.
- Wrap iwm_node casts into IWM_NODE() macro.
- Drop some fields:
* wt_hwqueue from Tx radiotap header;
* macaddr[6] from iwm_vap;
Approved by: adrian
Differential Revision: https://reviews.freebsd.org/D4753
melifaro [Sun, 3 Jan 2016 09:54:03 +0000 (09:54 +0000)]
Handle IPV6_PATHMTU option by spliting ip6_getpmtu_ctl() from ip6_getpmtu().
Add ro_mtu field to 'struct route' to be able to pass lookup MTU back to
the caller.
Currently, ip6_getpmtu() has 2 totally different use cases:
1) control plane (IPV6_PATHMTU req), where we just need to calculate MTU
and return it, w/o any reusability.
2) Actual ip6_output() data path where we (nearly) always use the provided
route lookup data. If this data is not 'valid' we need to perform another
lookup and save the result (which cannot be re-used by ip6_output()).
Given that, handle 1) by calling separate function doing rte lookup itself.
Resulting MTU is calculated by (newly-added) ip6_calcmtu() used by both
ip6_getpmtu_ctl() and ip6_getpmtu().
For 2) instead of storing ref'ed rte, store mtu (the only needed data
from the lookup result) inside newly-added ro_mtu field.
'struct route' was shrinked by 8(or 4 bytes) in r292978. Grow it again
by 4 bytes. New ro_mtu field will be used in other places like
ip/tcp_output (EMSGSIZE handling from output routines).
ngie [Sun, 3 Jan 2016 06:02:56 +0000 (06:02 +0000)]
- Use attach_md instead of hardcoding md(4) provider unit numbers
- Implement a gmirror_test_cleanup function, which in turn calls
geom_test_cleanup to clean up all md(4) providers allocated in the test
run.
- Remove duplicate logic in test scripts for removing md(4) providers.
- Don't create files in /tmp (outside the kyua sandbox); use the current
directory instead
imp [Sun, 3 Jan 2016 04:32:13 +0000 (04:32 +0000)]
If md_exec_hook is defined, provide a way to create the strings
for the environment variables we look up at runtime. Otherwise,
there's no way they will change, optimize it at compile time.
imp [Sun, 3 Jan 2016 04:32:04 +0000 (04:32 +0000)]
Add new LIBSOFT option. This is similar to the LIB32 option, except
for libraries that follow the soft float ABI. It's only supported on
armv6 as a transition to the new hard float ABI, so mark as broken
everywhere else.
ian [Sat, 2 Jan 2016 23:14:52 +0000 (23:14 +0000)]
Reword the comment to better describe what I found while researching the
problem that led to this temporary workaround (and also so I can properly
cite the PR in the commit this time).
In general this is intended to be a temporary workaround until we can figure
out why including any ram from the last megabyte of the physical address
space leads to a NULL pointer deref. Debugging that is made trickier by the
fact that I couldn't even get a backtrace in ddb.
andreast [Sat, 2 Jan 2016 22:04:37 +0000 (22:04 +0000)]
Fix booting of 32-bit kernels on 64-bit G5 hardware.
For rs6000, most memory insns and addi/addis do not allow GPR0 for RA
(they use literal zero there instead). So use a 'b' constraint to make
sure to have a base register other than GPR0.
GCC-4.7 and up handles this with allocating r9 instead of r0.
ian [Sat, 2 Jan 2016 22:00:52 +0000 (22:00 +0000)]
Use 64-bit math when processing the lists of physical and excluded memory
to generate the phys_avail and dump_avail arrays.
This is a partial fix for the kernel side of the problem mentioned in the
PR. This part handles the cases where comparing start and end addresses of
a block would fail because 32-bit wrap caused the end address to come out
zero if the end of the region is the end of the address space (0xffffffff
with 32-bit vm_paddr_t, but now the code should also work right if we ever
support LPAE with 36-bit addresses).
More work is necessary to make systems with ram at the end of the physical
address space usable, but at least initially it's going to be more like a
workaround than a fix, so this non-hacky part is being committed first.
kib [Sat, 2 Jan 2016 20:29:28 +0000 (20:29 +0000)]
Hide transient EBADF errors caused by the parallel revoke(2) or forced
unmount of devfs mounts, by restarting the failed syscall.
When restarted, failing syscalls eventually either stop finding the
node and returning ENOENT, or the vnode op vectors finally transition
to the deadfs vop. The later return EIO or other error, more
appropriate for the operation.
nwhitehorn [Sat, 2 Jan 2016 19:34:37 +0000 (19:34 +0000)]
Bump the maximum number of interrupt controllers to allow for the
proliferation of them on large IBM systems and add some error checking if
we exceed that number.
nwhitehorn [Sat, 2 Jan 2016 19:28:35 +0000 (19:28 +0000)]
Make using the #address-cells property on the interrupt parent in device
tree parsing opt-out rather than opt-in. All FDT-based systems as well as
PowerPC systems with real Open Firmware use the CHRP-derived binding that
includes it, which makes SPARC the odd man out here. Making it opt-out
avoids astonishment on new platform bring up.
ian [Sat, 2 Jan 2016 18:16:24 +0000 (18:16 +0000)]
Use 64-bit math when finding a block of ram to hold the kernel. This fixes
a problem on 32-bit systems which have ram occupying the end of the physical
address space -- for example, a block of ram at 0x80000000 with a size of
0x80000000 was overflowing 32 bit math and ending up with a calculated size
of zero.
This is a fix for one of the two problems mentioned in the PR. Something
similar will need to be done on the kernel side before the PR is closed.
adrian [Sat, 2 Jan 2016 17:14:22 +0000 (17:14 +0000)]
[ath] add explicit bus barriers.
The ath hal and driver code all assume the world is an x86 or the
bus layer does an explicit bus flush after each operation (eg netbsd.)
However, we don't do that.
So, to be "correct" on platforms like sparc64, mips and ppc (and maybe
ARM, I am not sure), just do explicit barriers after each operation.
Now, this does slow things down a tad on embedded platforms but I'd
rather things be "correct" versus "fast." At some later point if someone
wishes it to be fast then we should add the barrier calls to the HAL and
driver.
des [Sat, 2 Jan 2016 16:40:37 +0000 (16:40 +0000)]
Replace the cosine table with a sine table, which (due to the vagaries of
rounding) has better spread. Implement fp16_sin() to go along with
fp16_cos(). In the rendering loop, switch from addition to subtraction
so the center of the pattern will be a trough rather than a peak. This
is completely arbitrary, of course, but looks better to me.