Antonio Huete Jimenez [Mon, 30 Nov 2020 17:28:17 +0000 (18:28 +0100)]
libaura: More tests for dicts, duplicate handling
Antonio Huete Jimenez [Mon, 30 Nov 2020 14:42:11 +0000 (15:42 +0100)]
libaura: Add another test for dicts
Antonio Huete Jimenez [Mon, 30 Nov 2020 09:01:28 +0000 (10:01 +0100)]
libaura: Return NULL in some cases.
- Fail if there's not enough memory for allocating
num_buckets or if the method is unknown.
- Add test.
Antonio Huete Jimenez [Sat, 28 Nov 2020 07:37:20 +0000 (08:37 +0100)]
libaura: Staticize non-interface functions
Sepherosa Ziehau [Mon, 30 Nov 2020 14:36:31 +0000 (22:36 +0800)]
bce: Use IF_Mbps to avoid float number in kernel.
Reported-by: zrj
Sepherosa Ziehau [Wed, 4 Nov 2020 12:35:53 +0000 (20:35 +0800)]
re: Sync with Realtek's re-196.04; and fix re_start/re_tx_collect.
- Add several new chips' support, noticably RTL8125 support.
- Reorder chip related initializaton on the attach path a bit;
mainly to make sure ifnet is initialized.
- In re_start(), solely rely on the re_tx_free for the TX descriptor
setup loop. This avoids the fixup on the end of the loop. Add
assertion that the TX descriptor's corresponding mbuf is not setup
yet.
- On RTL8125, the EOF in the TX descriptor is no longer left uncleared
upon TX done, so simply rely on whether the TX descriptor's
corresponding mbuf is set or not to unload the busdma(9) and free the
TX mbuf. We actually should do this instead of rely on the EOF.
Tested-by: me (8101, 8111, 8168, 8125)
Romick <yellowrabbit2010 gmail.com> (8111)
Daniel Fojt <df neosystem.org> (8111)
Tomohiro Kusumi [Mon, 30 Nov 2020 15:06:34 +0000 (00:06 +0900)]
sys/vfs/hammer2: Make kprintf prefix consistent
Tomohiro Kusumi [Mon, 30 Nov 2020 15:07:40 +0000 (00:07 +0900)]
sys/vfs/hammer2: Don't ignore vn_isdisk() result on mount
Sascha Wildner [Sun, 29 Nov 2020 09:46:13 +0000 (10:46 +0100)]
nrelease: Define BUILDING_NRELEASE_DPORTS when building packages.
Requested by zrj to allow further optimization at the ports level.
Roy Marples [Sat, 28 Nov 2020 14:41:39 +0000 (14:41 +0000)]
Update dhcpcd README.DRAGONFLY
Roy Marples [Sat, 28 Nov 2020 14:38:03 +0000 (14:38 +0000)]
Merge branch 'vendor/DHCPCD'
Roy Marples [Sat, 28 Nov 2020 14:36:33 +0000 (14:36 +0000)]
Update to dhcpcd-9.3.4 with the following changes:
* DHCP: If error adding the address in oneshot, exit with failure
* DHCP: Only listen to the address if we successfully added it
* DHCP6: Fix segfault introduced in dhcpcd-9.3.3
* DHCP6: Abort in test mode when an error is returned by server
* options: allow --ia_na=1 and --ia_pd=2 on the command line
* options: Allow duid to take a value
Sascha Wildner [Sat, 28 Nov 2020 09:22:49 +0000 (10:22 +0100)]
<sys/sysproto.h>: Remove unneeded inclusion of <sys/sysmsg.h>.
After
80d831e1ad5c5886e45827bf13837cf84baba296, which removed the
struct sysmsg's in the *_args structures, this is no longer needed.
It also resolves circular #include issues because that commit at the
same time added #include <sys/sysproto.h> to <sys/sysmsg.h>.
Antonio Huete Jimenez [Fri, 27 Nov 2020 11:14:53 +0000 (12:14 +0100)]
sh(1) tests: dfregress support
Antonio Huete Jimenez [Fri, 27 Nov 2020 10:52:07 +0000 (11:52 +0100)]
dfregress: Add 'rc' option
- Allows to specify which return code will be considered success for the
test case being run. In the example below, a return code of '1' will
make the test PASS.
# Testcase type Options Args
builtins/alias.1 userland nobuild,rc=1 myarg1 myarg2
Antonio Huete Jimenez [Fri, 27 Nov 2020 10:09:31 +0000 (11:09 +0100)]
dfregress: Fix mistake in the previous commit
- We only return NULL on failure when fatal is false.
Antonio Huete Jimenez [Thu, 26 Nov 2020 22:40:00 +0000 (23:40 +0100)]
dfregress: Add 'interpreter' option
- Allows to specify the interpreter that will run the testcase file.
Some test cases just provide a file without execution permissions or no
shebang, etc. In some cases it is not desirable to modify the test case
files (i.e. contrib stuff).
Taking the example below, the previous behavior was to execvp() the file
'builtins/alias.0' but with 'interpreter' option, now you can specify
the shell you want to use to run it.
# Testcase type Options Args
builtins/alias.0 userland nobuild,interpreter=sh myarg1 myarg2
Sepherosa Ziehau [Thu, 26 Nov 2020 03:22:10 +0000 (11:22 +0800)]
udp: Port FreeBSD's IPPROTO_IP/IP_SENDSRCADDR.
Minor difference:
- Multicast src is not allowed.
- Always allow src, if it is same as inp_laddr and is not multicast.
While I'm here, fix datagram sending for multicast bound and connected
UDP socket.
Antonio Huete Jimenez [Wed, 25 Nov 2020 00:16:16 +0000 (01:16 +0100)]
tbridge: Add a section 4 man page.
Sascha Wildner [Tue, 24 Nov 2020 22:19:09 +0000 (23:19 +0100)]
<sys/poll.h>: Expose POLLSTANDARD and INFTIM only in the default env.
Cleans up the POSIX namespace.
Matthew Dillon [Tue, 24 Nov 2020 21:33:34 +0000 (13:33 -0800)]
vkernel - Fix the vkernel build after recent syscall changes (2)
* Another fix. Now the vkernel actually works again.
Reported-by: pikrzyszt, tuxillo
Matthew Dillon [Tue, 24 Nov 2020 19:57:20 +0000 (11:57 -0800)]
kernel - Fix sys_vmspace_ctl()
* This system call overwrites the trap-frame. The new syscall code
just has the argument pointer (uap) pointing into the trap-frame,
so in this particular case we have to copy the arguments out before
overwriting it.
Reported-by: pikrzyszt, tuxillo
Matthew Dillon [Tue, 24 Nov 2020 19:51:38 +0000 (11:51 -0800)]
kernel - Fix shared spin-lock starvation bug in VMs
* 'indefinite_uses_rdtsc' is set to zero by default when running in
a virtual machine (and set to 1 on a real machine). However, this
disables the windowing code in _spin_lock_shared_contested() and
causes it to defer to pending execusive lock requests indefinitely
under heavy-enough loads.
* Add a comment and always use a windowing test w/rdtsc() in
_spin_lock_shared_contested().
We were trying to avoid using the rdtsc() in VMs because some of them
apparently trap the rdtsc instruction. However, this puts us in a
no-win situation when it comes to dealing with spin-locks. So take the
hit and start using rdtsc again in some situations when operating in
a VM.
Reported-by: mjg
Sepherosa Ziehau [Tue, 24 Nov 2020 14:50:27 +0000 (22:50 +0800)]
inet: Refactor in_pcbbind() for IP_SENDSRCADDR support.
The newly factored out function will be used to validate address
passed through IP_SENDSRCADDR cmsg.
Sepherosa Ziehau [Tue, 24 Nov 2020 14:49:09 +0000 (22:49 +0800)]
tools/udpecho: Add command line option to prevent REUSEPORT.
Antonio Huete Jimenez [Tue, 24 Nov 2020 11:58:09 +0000 (12:58 +0100)]
tbridge: Add MODULE_DEPEND to the test module declaration.
- Makes sure you can't unload the tbridge module while
there are test modules loaded.
Sascha Wildner [Tue, 24 Nov 2020 09:18:05 +0000 (10:18 +0100)]
libc/stdio: Add missing MLINK.
Sepherosa Ziehau [Mon, 23 Nov 2020 15:36:52 +0000 (23:36 +0800)]
inpcb: Factor out function to obtain lport hash token.
Sepherosa Ziehau [Mon, 23 Nov 2020 15:02:13 +0000 (23:02 +0800)]
jail: Constify the address to be checked.
Sascha Wildner [Mon, 23 Nov 2020 08:42:35 +0000 (09:42 +0100)]
Put all definitions of htonl() and friends under _BYTEORDER_FUNC_DEFINED.
This is the intended way to properly deal with non-__GNUC__ cases (see
<machine/endian.h>). But it was only done properly in <arpa/inet.h> so
far.
Noticed-by: zrj
Sascha Wildner [Sun, 22 Nov 2020 08:27:36 +0000 (09:27 +0100)]
libc/stdio: Add _unlocked() flavors of fflush, fputc, fputs, fread, fwrite.
These are needed by some dports and can also be used in various things
in base. Bump __DragonFly_version, too.
Reported-by: zrj
Taken-from: FreeBSD
Sascha Wildner [Sun, 22 Nov 2020 08:14:08 +0000 (09:14 +0100)]
libc/siglist: Oops, CKPTX should be CKPTEXIT.
Sascha Wildner [Sun, 22 Nov 2020 07:04:38 +0000 (08:04 +0100)]
Clean up the namespace of <arpa/inet.h>, <netdb.h> and <netinet/in.h>.
These are further cleanups for stricter POSIX conformance, all related
to the uint*_t types, for which we previously had to include <stdint.h>
wholesale. Instead, wrap them in *_DECLARED macros and typedef them in
the headers that need them, usually just a subset.
Thanks to zrj for testing this with a dports bulk build.
Sascha Wildner [Sun, 22 Nov 2020 06:24:48 +0000 (07:24 +0100)]
sys/netproto/smb: Do not trust <netinet/in.h> to bring in <sys/types.h>.
<sys/types.h> will be removed from <netinet/in.h> in an upcoming commit.
Sascha Wildner [Sat, 21 Nov 2020 20:39:15 +0000 (21:39 +0100)]
Remove non-__STDC__ cases in a number of places.
We expect a standard C compiler for building DragonFly.
Sascha Wildner [Sat, 21 Nov 2020 20:35:20 +0000 (21:35 +0100)]
libc: Make the signal name array uppercase.
This is to satisfy POSIX which requires that the string printed by
'kill -l <signal number>' shall be the name of the symbolic constant
(for example, SIGKILL) without the 'SIG' part.
This also fixes a number of sh(1) tests.
While here, sync a bit further with FreeBSD. Specifically, use
C99 designated initialization.
Taken-from: FreeBSD
Sascha Wildner [Sat, 21 Nov 2020 19:34:07 +0000 (20:34 +0100)]
sh(1) tests: fix ld-elf.so version for DragonFly.
Sepherosa Ziehau [Sat, 21 Nov 2020 11:43:58 +0000 (19:43 +0800)]
test/netperf: Fix "clobbered" warning.
Sepherosa Ziehau [Sat, 21 Nov 2020 11:42:47 +0000 (19:42 +0800)]
Test commit.
Tomohiro Kusumi [Fri, 20 Nov 2020 16:27:42 +0000 (01:27 +0900)]
usr.sbin/fstyp: Fix HAMMER1 partial volumes detection (-u option)
Roy Marples [Fri, 20 Nov 2020 14:30:06 +0000 (14:30 +0000)]
dhcpcd: update README
Roy Marples [Fri, 20 Nov 2020 14:28:05 +0000 (14:28 +0000)]
Merge branch 'vendor/DHCPCD'
Roy Marples [Fri, 20 Nov 2020 14:26:27 +0000 (14:26 +0000)]
dhcpcd: Update to dhcpcd-9.3.3 with the following changes:
* dhcpcd: Don't create a launcher process if keeping in foreground
* dhcpcd: Add --noconfigure option
* control: Create an unpriv socket for non master mode
* options: Don't log unknown ones when printing pidfile location
Tomohiro Kusumi [Thu, 19 Nov 2020 16:10:05 +0000 (01:10 +0900)]
usr.sbin/fstyp: Properly support -l for HAMMER1
Tomohiro Kusumi [Wed, 18 Nov 2020 16:38:06 +0000 (01:38 +0900)]
usr.sbin/fstyp: Minor HAMMER1 fixes
Tomohiro Kusumi [Wed, 18 Nov 2020 15:46:41 +0000 (00:46 +0900)]
sbin/newfs_hammer2: Remove redundant \n in err/errx
Tomohiro Kusumi [Wed, 18 Nov 2020 15:35:33 +0000 (00:35 +0900)]
sbin/hammer2: Add bulkfree_tid for "volhdr"
Sascha Wildner [Wed, 18 Nov 2020 13:11:34 +0000 (14:11 +0100)]
<sys/callout.h>: Add a missing semicolon in the callout_set_arg() macro.
Fixes buildkernel with WANT_NETGRAPH7.
Antonio Huete Jimenez [Tue, 17 Nov 2020 19:37:53 +0000 (20:37 +0100)]
dsynth: Enable 'install' and 'deinstall' phases on test.
Antonio Huete Jimenez [Tue, 17 Nov 2020 12:12:43 +0000 (13:12 +0100)]
dsynth: Remove any trace of 'install-mtree' phase.
- It's an empty target, poudriere-specific and it will
be removed from ports at some point.
Sascha Wildner [Tue, 17 Nov 2020 12:04:00 +0000 (13:04 +0100)]
Update the pciconf(8) database.
November 14, 2020 snapshot from https://pci-ids.ucw.cz
Matthew Dillon [Mon, 16 Nov 2020 10:09:38 +0000 (02:09 -0800)]
kernel - Rewrite the callout API, major structural changes (2)
* Fix deadlock during recursive callback, triggered by the USB
subsystem.
Basically we have to clear the SET bit immediately after setting
INPROG, and also allow normal processing to occur synchronously
from the same thread that dispatched the callback (normally INPROG
prevents normal processing from occuring).
Reported-by: zrj
Matthew Dillon [Sun, 15 Nov 2020 19:41:26 +0000 (11:41 -0800)]
kernel - Add PROC_PDEATHSIG_CTL and PROC_PDEATHSIG_STATUS
* Add PROC_PDEATHSIG_CTL and PROC_PDEATHSIG_STATUS to procctl(2).
This follows the linux and freebsd semantics, however it should be noted
that since the child of a fork() clears the setting, these semantics have
a fork/exit race between an exiting parent and a child which has not
yet setup its death wish.
* Also fix a number of signal ranging checks.
Requested-by: zrj
Matthew Dillon [Fri, 13 Nov 2020 22:38:04 +0000 (14:38 -0800)]
kernel - Rewrite the callout API, major structural changes
* Make struct callout much smaller, primarily now just containing
a pointer to a shadow structure which is allocated on first use.
The callout backend uses this shadow structure on the callout
lists instead of the embedded struct callout. This should reduce
occurances of untraceable panics due to list corruption. Instead,
the callout-v-shadow structure linkages are validated with an assertion
and if this panics it should be possible to determine which subsystem
was responsible from the kernel core.
* tsleep() etc is used early in the boot sequence and use some
bypass functions to streamline operations and avoid allocation
calls. This should also improve tsleep()s performance a bit over
the original implementation.
Sascha Wildner [Sun, 15 Nov 2020 01:36:51 +0000 (02:36 +0100)]
Sync ACPICA with Intel's version
20201113.
* Fix a memory leak that could occur in _HID repair.
* Improve namespace debugging.
* Handle "orphan" _REG methods.
* Improve iasl(8) and acpihelp(8).
* Add 5 new NVDIMM and NFIT table related UUIDs to the known UUID table.
For detailed list, please see sys/contrib/dev/acpica/changes.txt.
François Tigeot [Fri, 13 Nov 2020 13:34:59 +0000 (14:34 +0100)]
drm/linux: Add llist_for_each_entry() and llist_del_first()
Obtained-from: OpenBSD
François Tigeot [Fri, 13 Nov 2020 13:17:40 +0000 (14:17 +0100)]
kernel: Add atomic_cas_ptr()
Obtained-from: OpenBSD
François Tigeot [Wed, 11 Nov 2020 19:49:32 +0000 (20:49 +0100)]
drm/linux: Rewrite the workqueue implementation
Most workqueues now use one runner thread per logical CPU, except for
WQ_UNBOUND ones, which only have a single kernel thread.
This commit contains critical fixes from Matthew Dillon for multithread
interaction issues in queue_work() and the cancel functions.
François Tigeot [Tue, 10 Nov 2020 13:58:11 +0000 (14:58 +0100)]
drm/linux: Add atomic_dec_if_positive()
Obtained-from: FreeBSD
Sascha Wildner [Tue, 10 Nov 2020 10:03:28 +0000 (11:03 +0100)]
libdialog: Add missing function attribute definitions. Raise WARNS to 6.
For some reason, 'configure' does not define them ootb.
Sascha Wildner [Mon, 9 Nov 2020 15:59:58 +0000 (16:59 +0100)]
kernel/ufs: Add some __dead2 to the stubs for the non-SOFTUPDATES case.
Matthew Dillon [Mon, 9 Nov 2020 01:25:03 +0000 (17:25 -0800)]
build - Adjust message output after a make build-all
* Clarify that the recommended pkg upgrade -f should be run after
rebooting into the new system.
Matthew Dillon [Mon, 9 Nov 2020 01:24:09 +0000 (17:24 -0800)]
dmesg - Check current directory when -n is used
* When -n is used, first check the current directory for kern.n and
vmcore.n. If either is not found, then check /var/crash as before.
Sascha Wildner [Sat, 7 Nov 2020 23:18:43 +0000 (00:18 +0100)]
kernel: Use howmany() in a couple of places.
Matthew Dillon [Sat, 7 Nov 2020 18:42:33 +0000 (10:42 -0800)]
kernel - Incorporate drm79.patch to master, cleaned up
* Should fix a number of niggling issues in master now,
that are already incorporated into the next drm sync.
Matthew Dillon [Sat, 7 Nov 2020 18:19:17 +0000 (10:19 -0800)]
kernel - Change pager interface to pass page index 3/*
* Don't shortcut vm_object_page_remove() with a resident_page_count
test for MGTDEVICE objects. These objects are not required to have
their VM pages entered into them.
* in vm_object_page_remove(), change pmap_remove_pages() to
pmap_remove(). The former is meant to be used only in the
exit code and does not bother with TLB synchronization.
pmap_remove() properly handles any TLB synchronization.
Sascha Wildner [Fri, 6 Nov 2020 16:13:14 +0000 (17:13 +0100)]
Add a minimal kernel configuration file.
For testing the kernel build without any options.
Sascha Wildner [Fri, 6 Nov 2020 16:10:05 +0000 (17:10 +0100)]
LINT64: Note that 'ident' and 'cpu' are mandatory, too.
Sascha Wildner [Fri, 6 Nov 2020 15:06:40 +0000 (16:06 +0100)]
kernel: Fix building without 'options _KPOSIX_PRIORITY_SCHEDULING'.
Prototype generation for this case was broken after
80d831e1ad5c5886e458.
Reported-by: zrj
Daniel Fojt [Wed, 4 Nov 2020 20:52:19 +0000 (21:52 +0100)]
Upgrade libressl. 2/2
Update README.{DELETED,DRAGONFLY} and adapt Makefiles.
Daniel Fojt [Wed, 4 Nov 2020 20:19:57 +0000 (21:19 +0100)]
Upgrade libressl. 1/2
Merge branch 'vendor/LIBRESSL' into master.
Daniel Fojt [Tue, 3 Nov 2020 18:37:17 +0000 (19:37 +0100)]
vendor/libressl: upgrade from 3.1.4 to 3.2.2
Summary of notable changes:
* new TLSv1.3 implementation enabled by default for both client
and server
* new X509 certificate chain validator that correctly handles
multiple paths through intermediate certificates
* new name constraints verification implementation which passes
the bettertls.com certificate validation check suite
* numerous leaks and out-of-bounds write fixed
* many improvements, refactoring, optimizations and code cleanups
See ChangeLog for details.
Sascha Wildner [Thu, 5 Nov 2020 05:04:55 +0000 (06:04 +0100)]
kernel: Reindent a few functions that were not using tabs.
While here, fix a typo in the comments.
Sascha Wildner [Thu, 5 Nov 2020 03:13:11 +0000 (04:13 +0100)]
kernel/amd{smn,temp}: Sync with FreeBSD.
Adds support for Ryzen 4000 APU (Zen 2, "Renoir").
Reported-by: zrj
zrj [Wed, 4 Nov 2020 10:31:21 +0000 (10:31 +0000)]
lib/csu: Add ifunc support for statically linked programs.
This assumes that both _init_tls() and _rtld_call_init() paths in libc
will not contain ifunc calls. For now should be enough since there was
no ifunc support previously.
Derived-from: NetBSD crt0-common.c implementation
zrj [Tue, 3 Nov 2020 17:28:17 +0000 (17:28 +0000)]
rtld: Handle non-PLT type 37 relocations.
Mainly to unbreak certain use cases of ifunc.
zrj [Tue, 3 Nov 2020 17:22:11 +0000 (17:22 +0000)]
rtld: Extract reloc_iresolve_one().
Will be used to properly handle non-PLT 37 ifunc relocations in -pie.
Matthew Dillon [Wed, 4 Nov 2020 21:51:40 +0000 (13:51 -0800)]
kernel - Change pager interface to pass page index 2/2
* Adjust the DRM calls to vm_pager_get_page() to pass the
page index.
* Greatly simplify drm/linux_shmem.c and drm/ttm/ttm_bo_vm.c,
removing the need to deal with placemarker pages for
VM faults on OBJT_MGTDEVICE objects.
Matthew Dillon [Wed, 4 Nov 2020 21:49:03 +0000 (13:49 -0800)]
kernel - Change pager interface to pass page index 1/2
* Change the *getpage() API to include the page index as
an argument. This allows us to avoid passing any vm_page_t
for OBJT_MGTDEVICE VM pages.
By removing this requirement, the VM system no longer has to
pre-allocate a placemarker page for DRM faults and the DRM
system can directly install the page in the pmap without
tracking it via a vm_page_t.
Matthew Dillon [Wed, 4 Nov 2020 21:34:09 +0000 (13:34 -0800)]
kernel - Fix uninitialized variable in kqueue_register()
* Fix another uninitialized variable that gcc didn't detect,
this time in kqueue_register().
* Caused kernel compiled with -O0 to not operate properly.
* Initializing the error variable to 0 solves the problem.
Reported-by: zrj
Matthew Dillon [Wed, 4 Nov 2020 18:35:22 +0000 (10:35 -0800)]
kernel - Fix missing initialization in vnode_pager_lock()
* struct vnode *vp was not being initialized to NULL, causing
a garbage return not caught by gcc if no while() condition
is met.
* Only occurred with -O0. The optimized case doesn't seem to hit
the issue through luck.
Reported-by: zrj
Tomohiro Kusumi [Tue, 3 Nov 2020 16:30:19 +0000 (01:30 +0900)]
sys/vfs/hammer2: Illustrate level1-6 freemap
zrj [Mon, 2 Nov 2020 18:38:20 +0000 (18:38 +0000)]
kernel: Avoid spurious diagnostic with older gcc47.
zrj [Mon, 2 Nov 2020 18:34:45 +0000 (18:34 +0000)]
bsd.sys.mk: Make -Wmissing-field-initializers non-fatal.
It was decided that gcc47 gives few diagnostics wrongly in C99 mode.
Sascha Wildner [Sun, 1 Nov 2020 20:45:06 +0000 (21:45 +0100)]
Sync zoneinfo database with tzdata2020d from ftp://ftp.iana.org/tz/releases
* Palestine ends DST earlier than predicted, on 2020-10-24.
* Fiji starts DST later than usual, on 2020-12-20.
For a detailed list of changes, see share/zoneinfo/NEWS.
Sascha Wildner [Sun, 1 Nov 2020 19:59:43 +0000 (20:59 +0100)]
script.7: Fix description of the length of the '#!' line.
Sascha Wildner [Sun, 1 Nov 2020 19:58:49 +0000 (20:58 +0100)]
environ.7: Sync with FreeBSD.
Daniel Fojt [Sun, 1 Nov 2020 16:53:52 +0000 (17:53 +0100)]
kernel: fix getcwd(3) return value for non-existing directory
In case current directory no longer exists, properly return ENOENT
from getcwd(), as described in manpage.
Issue: https://bugs.dragonflybsd.org/issues/3250
Tomohiro Kusumi [Sun, 1 Nov 2020 15:10:09 +0000 (00:10 +0900)]
sys/vfs/hammer2: Remove obsolete comment on copyids
copyids in inode was removed in
ae183399278ee71fba31527cd18a7545ac9af2e4 in 2012.
Roy Marples [Sun, 1 Nov 2020 14:37:48 +0000 (14:37 +0000)]
Note dhcpcd update
Roy Marples [Sun, 1 Nov 2020 14:35:46 +0000 (14:35 +0000)]
Merge branch 'vendor/DHCPCD'
Roy Marples [Sun, 1 Nov 2020 14:33:08 +0000 (14:33 +0000)]
Update to dhcpcd-9.3.2 with the following changes:
* DHCP: Add support for IPv6-Only Preferred option, RFC 8925.
* BSD: `LINK_STATE_UNKNOWN` is treated as UP once again
* privsep: pass logging to the privileged actioneer
* privsep: allow logfile re-opening to work
* privsep: close BPF socket on ENXIO
* privsep: don't leave a BOOTP BPF listener rebooting in non master mode
zrj [Sat, 31 Oct 2020 19:19:42 +0000 (19:19 +0000)]
boot/pc32: Fix ambiguity in legacy loaders for Hammer selection.
Explicitly name both variants. Default Hammer2 stays in legacy loader.
Allow to built loader with HAMMER(5) support for testing purposes.
Adjust libstand(3).
zrj [Sat, 31 Oct 2020 18:57:55 +0000 (18:57 +0000)]
boot/pc32: Reorder fs probe list.
Sort by: disk, cd, net, file.
This should increase lookup speed for network loaders.
zrj [Sat, 31 Oct 2020 18:11:03 +0000 (18:11 +0000)]
boot/pc32: Conditionalize splitfs support in legacy loaders.
This floppy feature was inherited and not so useful anymore. Disabling
it reduces filename lookups by half. Also put it last, if enabled.
zrj [Sat, 31 Oct 2020 09:12:38 +0000 (09:12 +0000)]
Switch to ld.bfd as default ld for world stage.
Until boot/efi issues are resolved:
* ld.gold not respecting linker script for section merges;
* clang built boot1.efi and loader.efi failures.
Default can be restored by setting WORLD_LDVER=ld.gold in make.conf(5).
Final installed ld(1) will still default to ld.gold to not disturb
things in DPorts.
NOTE: this might produce different final binaries if building in tree,
but it should not be fatal as we support both ld.bfd and ld.gold.
Use "make LDVER=ld.bfd" to mimic buildworld behaviour when needed.
zrj [Sat, 31 Oct 2020 08:31:02 +0000 (08:31 +0000)]
Switch to binutils234 as default base binutils.
zrj [Sat, 31 Oct 2020 08:29:40 +0000 (08:29 +0000)]
boot/efi: Objcopy more sections to final *.efi objects.
Unbreaks uefi boot with when compiled with ld.gold from binutils234.
zrj [Sat, 31 Oct 2020 06:44:42 +0000 (06:44 +0000)]
boot/pc32: Consistently disable integrated-as for intermediates.
This ensures that produced assembly intermediates are compatible with
external as(1) when compiling with clang.
zrj [Sat, 31 Oct 2020 06:42:39 +0000 (06:42 +0000)]
kernel: Disable use of clang's integrated as.