dim [Sat, 5 Nov 2011 21:16:39 +0000 (21:16 +0000)]
Make it possible to set CC and CXX (and optionally, AS and LD) in
make.conf(5), while allowing the build32 stage on 64-bit architectures
to still override them, so that stage can successfully build 32-bit
compatibility libraries.
Explanation:
1) The build32 stage sets environment variables CC, CXX, AS and LD for
its sub-make, to add 32-bit specific flags (-m32 and such).
2) The sub-make reads sys.mk, encounters CC?= and CXX?= assignments, so
does not alter them.
3) After some other stuff, sys.mk reads /etc/make.conf. When you have
"CC=xxx" and "CXX=yyy" statements in there, they will *override* the
build32-supplied CC/CXX values, nullifying the 32-bit specific flags.
4) Thus all objects get built as 64-bit anyway, and since LD is usually
not set in make.conf, it still has the 32-bit flags!
5) Now, whenever something is linked, you will get a "ld: Relocatable
linking with relocations from format elf64-x86-64-freebsd (foo.o) to
format elf32-i386-freebsd (bar.o) is not supported" error.
Fix this by adding "-ECC -ECXX -EAS -ELD" to the build32 sub-make
invocation, which forces those environment variables to always override
any assignment in makefiles. Thus making it possible to simply set:
jmg [Sat, 5 Nov 2011 17:55:49 +0000 (17:55 +0000)]
group functions together by function...
document knlist_delete, and better document what knlist_clear does... Note
that both of these functions may sleep, and also unlock/relock the list
lock...
dim [Sat, 5 Nov 2011 16:47:47 +0000 (16:47 +0000)]
Whenever you boot with nfsv4_server_enable=NO (the default) in rc.conf,
the /etc/rc.d/nfsd script sets vfs.nfsd.server_max_nfsvers to 3.
Then, when you set nfsv4_server_enable=YES in rc.conf, and restart nfsd
via the rc.d script, without rebooting, the sysctl does *not* get reset
to max version 4, so NFSv4 still doesn't work.
Fix this by explicitly setting vfs.nfsd.server_max_nfsvers to 4 when
NFSv4 is requested.
I also added resetting of the nfs_privport sysctls, since this has the
same issue: nfs_reserved_port_only=YES in rc.conf sets the nfs_privport
sysctl to 1, but in the other case, the sysctl doesn't get reset to 0.
Reviewed by: rmacklem
Silence from: rc@
MFC after: 3 days
pjd [Sat, 5 Nov 2011 16:29:03 +0000 (16:29 +0000)]
In zvol_open() if the spa_namespace_lock is already held, it means that
ZFS is trying to open and taste ZVOL as its VDEV. This is not supported,
so return an error instead of panicing on spa_namespace_lock recursion.
Reported by: Robert Millan <rmh@debian.org>
PR: kern/162008
MFC after: 3 days
This variable was added in r82352 back in 2001, but even then it didn't
have any use. Because it's not marked static, the C compiler won't
complain about it.
kib [Sat, 5 Nov 2011 08:20:32 +0000 (08:20 +0000)]
Provide typedefs for the type of bit mask for the page bits.
Use the defined types instead of int when manipulating masks.
Supposedly, it could fix support for 32KB page size in the
machine-independend VM layer.
yongari [Fri, 4 Nov 2011 22:53:52 +0000 (22:53 +0000)]
Don't abuse if_hwassist and make sure enabling corresponding TX/RX
checksum offloading and VLAN hardware tag insertion/stripping from
the currently enabled hardware offloading capabilities.
Previously if_hwassist, which was initialized to TX/RX checksum
offloading, was blindly used to enable both TX and RX checksum
offloading such that disabling either TX or RX checksum offloading
was not possible.
ti(4) controllers support TX/RX checksum offloading with VLAN
tagging so announce TX/RX checksum offloading capability over VLAN
to vlan(4).
Make VLAN hardware tag insertion/stripping honors currently enabled
interface capability instead of blindly enabling VLAN hardware
tagging. This change allows disabling hardware support of VLAN tag.
Because ti(4) supports VLAN oversized frames, make network stack
know the capability by setting if_hdrlen.
While I'm here, rewrite SIOCSIFCAP handler and make sure to
reinitialize controller whenever TX/RX checksum offloading and VLAN
hardware tagging option is changed. The requirement of controller
reinitialization comes from the limitation of Tigon I/II firmware.
Tigon I/II firmware requires all related RCBs should be
reinitialized whenever any of its hardware offloading capabilities
change.
vlan(4) is also notified whenever the parent interface's capability
changes such that it can correctly handle TX/RX checksum offloading
based on parent interface's enabled offloading capabilities.
RX checksum offloading handler was changed to make upper stack use
controller computed partial checksum value. Previously, ti(4) just
set the computed value for any frames(IPv4, IPv6) and the value was
not used in upper stack because driver didn't set CSUM_DATA_VALID
such that upper network stack had to recompute checksum of TCP/UDP
packets. I have no idea how this was not noticed for a long time.
With this change, upper network stack does not have to fully
recompute the checksum such that calculating pseudo checksum based
on partial checksum is sufficient to know whether received packet's
checksum is correct or not. However, I don't know why ti(4) does
not have controller compute pseudo checksum as controller has
ability to do it. I'm just guessing enabling that feature could
trigger a firmware bug or could be slower than computing it on host
side so just leave it as it was.
In order not to produce false positives, ti(4) now checks whether
controller actually computed IP or TCP/UDP checksum by checking
ti_flags field.
yongari [Fri, 4 Nov 2011 21:42:13 +0000 (21:42 +0000)]
Don't clear upper 4bits from VLAN tag information. It's
responsibility of vlan(4) to extract VLAN id from the tag
information and vlan(4) correctly handles it.
yongari [Fri, 4 Nov 2011 20:43:37 +0000 (20:43 +0000)]
Announce IFCAP_LINKSTATE capability and let network stack know link
state changes. Hide superfluous link up/down message under
bootverbose since if_link_state_change(9) shows that information.
While I'm here, change baudrate with the resolved speed of the
established link instead of blindly setting it 1G. Unfortunately,
it seems there is no way to differentiate 10/100Mbps from
non-gigabit link so just assume we established a 100Mbps link if
current link is not a gigabit link.
The size passed to strlcat() must depend on the input length, not the
output length. Because the input and output buffers are equal in size,
the resulting binary does not change at all.
yongari [Fri, 4 Nov 2011 18:39:39 +0000 (18:39 +0000)]
Make ti(4) build with 'options TI_PRIVATE_JUMBOS'.
This was broken in r175872.
We have a UMA backed jumbo allocator and that is much better
implementation than having a local jumbo buffer allocator in
driver. This local allocator would be removed in near future but
fixing build before removal wouldn't be a bad idea.
Always use the opt_*.h options for ipfw.ko, not just when
compiled into the kernel.
Do not try to build the module in case of no INET support but
keep #error calls for now in case we would compile it into the
kernel.
This should fix an issue where the module would fail to enable
IPv6 support from the rc framework, but also other INET and INET6
parts being silently compiled out without giving a warning in the
module case.
While here garbage collect unneeded opt_*.h includes.
opt_ipdn.h is not used anywhere but we need to leave the DUMMYNET
entry in options for conditional inclusion in kernel so keep the
file with the same name.
Add missing static keywords for global variables to tools in sbin/.
These tools declare global variables without using the static keyword,
even though their use is limited to a single C-file, or without placing
an extern declaration of them in the proper header file.
jhb [Fri, 4 Nov 2011 04:02:50 +0000 (04:02 +0000)]
Add the posix_fadvise(2) system call. It is somewhat similar to
madvise(2) except that it operates on a file descriptor instead of a
memory region. It is currently only supported on regular files.
Just as with madvise(2), the advice given to posix_fadvise(2) can be
divided into two types. The first type provide hints about data access
patterns and are used in the file read and write routines to modify the
I/O flags passed down to VOP_READ() and VOP_WRITE(). These modes are
thus filesystem independent. Note that to ease implementation (and
since this API is only advisory anyway), only a single non-normal
range is allowed per file descriptor.
The second type of hints are used to hint to the OS that data will or
will not be used. These hints are implemented via a new VOP_ADVISE().
A default implementation is provided which does nothing for the WILLNEED
request and attempts to move any clean pages to the cache page queue for
the DONTNEED request. This latter case required two other changes.
First, a new V_CLEANONLY flag was added to vinvalbuf(). This requests
vinvalbuf() to only flush clean buffers for the vnode from the buffer
cache and to not remove any backing pages from the vnode. This is
used to ensure clean pages are not wired into the buffer cache before
attempting to move them to the cache page queue. The second change adds
a new vm_object_page_cache() method. This method is somewhat similar to
vm_object_page_remove() except that instead of freeing each page in the
specified range, it attempts to move clean pages to the cache queue if
possible.
To preserve the ABI of struct file, the f_cdevpriv pointer is now reused
in a union to point to the currently active advice region if one is
present for regular files.
jhb [Fri, 4 Nov 2011 03:39:31 +0000 (03:39 +0000)]
Move the cleanup of f_cdevpriv when the reference count of a devfs
file descriptor drops to zero out of _fdrop() and into devfs_close_f()
as it is only relevant for devfs file descriptors.
rmacklem [Thu, 3 Nov 2011 14:38:03 +0000 (14:38 +0000)]
Both a crash reported on freebsd-current on Oct. 18 under the
subject heading "mtx_lock() of destroyed mutex on NFS" and
PR# 156168 appear to be caused by clnt_dg_destroy() closing
down the socket prematurely. When to close down the socket
is controlled by a reference count (cs_refs), but clnt_dg_create()
checks for sb_upcall being non-NULL to decide if a new socket
is needed. I believe the crashes were caused by the following race:
clnt_dg_destroy() finds cs_refs == 0 and decides to delete socket
clnt_dg_destroy() then loses race with clnt_dg_create() for
acquisition of the SOCKBUF_LOCK()
clnt_dg_create() finds sb_upcall != NULL and increments cs_refs to 1
clnt_dg_destroy() then acquires SOCKBUF_LOCK(), sets sb_upcall to
NULL and destroys socket
This patch fixes the above race by changing clnt_dg_destroy() so
that it acquires SOCKBUF_LOCK() before testing cs_refs.
attilio [Thu, 3 Nov 2011 14:36:56 +0000 (14:36 +0000)]
Disable interrupt and preemption for smp_rendezvous() also in the
UP/!SMP case.
The callbacks may be relying on this feature and having 2 different
ways to deal with them is not correct.
kib [Wed, 2 Nov 2011 18:08:30 +0000 (18:08 +0000)]
Despite official i386 ABI does not mandate any stack alignment besides
the word alignment, some versions of gcc do require 16-byte alignment.
Make sure the stack is 16-byte aligned before calling a subroutine.
kib [Wed, 2 Nov 2011 18:06:22 +0000 (18:06 +0000)]
Make sure that stack is 16-byte aligned before calling a function,
as it is required by amd64 ABI. Add a comment for the places were
the stack is accidentally properly aligned already.
mav [Wed, 2 Nov 2011 09:24:59 +0000 (09:24 +0000)]
Add mutex and two flags to make orphan() call properly asynchronous:
- delay consumer closing and detaching on orphan() until all I/Os complete;
- prevent new I/Os submission after orphan() called.
Previous implementation could destroy consumers still having active
requests and worked only because of global workaround made on GEOM level.
dougb [Wed, 2 Nov 2011 07:40:23 +0000 (07:40 +0000)]
1. Add a new option, --run-updates, to always or never run the newalises,
pwd_mkdb, etc. updates at the end of the comparison.
2. Add an update to the end to handle /etc/localtime, if it exists.
If /var/db/zoneinfo exists, automatically update /etc/localtime,
which should (hopefully) be safe to do. If not, prompt the user
to run tzsetup.
3. Update run_it_now(), the function that handles input for the updates,
to make sure that we got a valid answer, and to handle the --run-updates
option if supplied.
dougb [Wed, 2 Nov 2011 04:21:20 +0000 (04:21 +0000)]
If the user is moving from any other time zone to UTC we need
to delete any old /var/db/zoneinfo file that may exist so that
tzsetup -r does the right thing.
mav [Tue, 1 Nov 2011 23:12:22 +0000 (23:12 +0000)]
Make orphan() method in geom_dev asynchronous using destroy_dev_sched_cb()
instead of destroy_dev(). It moves device destruction waiting out of the
topology lock and so fixes dead lock between orphanization and closing.
Real provider and geom destruction called from swi context after device
destroyed as callback of the destroy_dev_sched_cb().
marius [Tue, 1 Nov 2011 21:26:57 +0000 (21:26 +0000)]
Add a PCI front-end to esp(4) allowing it to support AMD Am53C974 and
replace amd(4) with the former in the amd64, i386 and pc98 GENERIC kernel
configuration files. Besides duplicating functionality, amd(4), which
previously also supported the AMD Am53C974, unlike esp(4) is no longer
maintained and has accumulated enough bit rot over time to always cause
a panic during boot as long as at least one target is attached to it
(see PR 124667).
mav [Tue, 1 Nov 2011 20:56:19 +0000 (20:56 +0000)]
Refactor disk disconnection and geom destruction handling sequences.
Do not close/destroy opened consumer directly in case of disconnect. Instead
keep it existing until it will be closed in regular way in response to
upstream provider destruction. Delay geom destruction in the same way.
Previous implementation could destroy consumers still having active
requests and worked only because of global workaround made on GEOM level.
marius [Tue, 1 Nov 2011 18:28:33 +0000 (18:28 +0000)]
Increase the IOC port initialization timeouts by ten times to what the
corresponding Linux driver uses. This allows mpt(4) to still recognize
all good SATA devices in presence of a defective one, which takes about
45 seconds.
In the long term we probably should implement the logic used by mpt2sas(4)
allowing IOC port initialization to complete at a later time.
marius [Tue, 1 Nov 2011 17:57:21 +0000 (17:57 +0000)]
In r225931 I've missed the only other driver using the pointer returned
by rman_get_virtual(9) to access device registers sparc64 currently cares
about.
Ideally ata(4) should just be converted to access these using bus_space(9)
read/write functions instead as there's really no reason to do it the
former way. However, this part of ata-siliconimage.c should go away in
favor of siis(4) sooner or later anyway and I don't have the hardware to
actually test the SX4 bits of ata-promise.c.
Also ideally the other architectures should also properly handle the
BUS_SPACE_MAP_LINEAR flag of bus_space_map(9) so this code wouldn't need
to be #ifdef'ed.
mav [Tue, 1 Nov 2011 17:04:42 +0000 (17:04 +0000)]
Refactor disk disconnection and geom destruction handling sequences.
Do not close/destroy opened consumer directly in case of disconnect. Instead
keep it existing until it will be closed in regular way in response to
upstream provider destruction. Delay geom destruction in the same way.
Previous implementation could destroy consumers still having active
requests and worked only because of global workaround made on GEOM level.
marius [Tue, 1 Nov 2011 16:13:59 +0000 (16:13 +0000)]
- Import the common MII bitbang'ing code from NetBSD and convert drivers to
take advantage of it instead of duplicating it. This reduces the size of
the i386 GENERIC kernel by about 4k. The only potential in-tree user left
unconverted is xe(4), which generally should be changed to use miibus(4)
instead of implementing PHY handling on its own, as otherwise it makes not
much sense to add a dependency on miibus(4)/mii_bitbang(4) to xe(4) just
for the MII bitbang'ing code. The common MII bitbang'ing code also is
useful in the embedded space for using GPIO pins to implement MII access.
- Based on lessons learnt with dc(4) (see r185750), add bus barriers to the
MII bitbang read and write functions of the other drivers converted in
order to ensure the intended ordering. Given that register access via an
index register as well as register bank/window switching is subject to the
same problem, also add bus barriers to the respective functions of smc(4),
tl(4) and xl(4).
- Sprinkle some const.
Thanks to the following testers:
Andrew Bliznak (nge(4)), nwhitehorn@ (bm(4)), yongari@ (sis(4) and ste(4))
Thanks to Hans-Joerg Sirtl for supplying hardware to test stge(4).
Reviewed by: yongari (subset of drivers)
Obtained from: NetBSD (partially)
mav [Tue, 1 Nov 2011 08:57:49 +0000 (08:57 +0000)]
Workaround the problem introduced by combination of r162200 and r215687.
r162200 delays provider orphanization until all running requests complete,
to workaround broken orphan() method implementation in some classes.
r215687 removes persistent periodic (10Hz) event thread wake ups.
Together these changes can indefinitely delay orphanization until some
other event wake up the event thread. One consequence of this is inability
of CAM to destroy device disconnected when busy and, as consequence, create
new one after reconnection.
While the best solution would be to revert r162200, it is not easy, as
some classes still look broken in that way. Instead conditionally wake up
event thread if there are some providers waiting for orphanization.
ed [Mon, 31 Oct 2011 08:59:17 +0000 (08:59 +0000)]
Put some static keywords in the source code.
For these simple utilities, it doesn't harm to make all global variables
static. In fact, this allows the compiler to perform better forms of
optimisation and analysis.
dim [Sun, 30 Oct 2011 22:20:17 +0000 (22:20 +0000)]
Pull in r143305 and r143312 from upstream clang trunk, so using "clang
-march=native" on AMD K10 family processors no longer errors out with
"unknown target CPU 'amdfam10'". This also enables use of SSE4A.
Reported by: David Marec <david.marec@davenulle.org>
MFC after: 3 days
marius [Sun, 30 Oct 2011 21:17:42 +0000 (21:17 +0000)]
- Use device_t rather than the NetBSDish struct device.
- Move esp_devclass to ncr53c9x.c in order to allow different bus front-ends
to use it.
- Use KOBJMETHOD_END.
- Remove the gl_clear_latched_intr hook as it's not needed for any of the
chips nor the front-ends supported in FreeBSD and likely never will be.
- Correct the DMA constraints used in the SBus front-end, the LSI64854 isn't
limited to 32-bit DMA.
- The ESP200 also only supports up to 64k transfers.
- Don't let the DMA and SBus front-end supply a maximum transfer size larger
than MAXPHYS as that's the maximum the upper layers use and we otherwise
just waste resources unnecessarily.
- Initialize the ECB callout and don't zero the handle when returning ECBs
to the free list so that ncr53c9x_callout() actually is called with the
driver lock held.
- On detach the driver lock should be held across cam_sim_free() according
to isp(4) and a panic received.
- Check the return value of NCRDMA_SETUP(), i.e. bus_dmamap_load(9), and try
to handle failures gracefully.
- In ncr53c9x_action() replace N calls to xpt_done() in a switch with just
one at the end.
- On XPT_PATH_INQ report "NCR" rather than "Sun" as the vendor as the former
is somewhat more correct as well as the maximum supported transfer size via
maxio in order to take advantage of controllers that that can handle more
than DFLTPHYS.
- Print the number of MESSAGE (EXTENDED) rejected.
- Fix the path encoded in the multiple inclusion protection of ncr53c9xvar.h.
- Correct the DMA constraints used in the LSI64854 core to not exceed the
maximum supported transfer size and include the boundary so we don't need
to check on every setup of a DMA transfer.
- Let the bus DMA map callbacks do nothing in case of an error.
- Correctly handle > 64k transfers for FAS366 in the LSI64854. A new feature
flag NCR_F_LARGEXFER was introduced so we just need to check for this one
and not for individual controllers supporting large transfers in several
places.
- Let the LSI64854 core load transfer buffers using BUS_DMA_NOWAIT as the
NCR53C9x core can't handle EINPROGRESS. Due to lack of bounce buffers
support, sparc64 doesn't actually use EINPROGRESS and likely never will,
as an example for writing additional front-ends for the NCR53C9x core it
makes sense to set BUS_DMA_NOWAIT anyway though.
- Some minor cleanup.
rmh [Sun, 30 Oct 2011 16:29:04 +0000 (16:29 +0000)]
Fix a few gratuitous library dependencies. Some of the ZFS utilities
are linked with libraries they don't use:
- zinject doesn't use libavl
- ztest doesn't use libz
- zdb uses neither libavl nor libz
- zfs uses neither libbsdxml nor libm, nor libsbuf
- zpool uses neither libbsdxml nor libm, nor libsbuf
In addition, libzfs needs libm because it uses pow(), however it isn't
linked with -lm. This went unnoticed because all its users had -lm before.
alc [Sun, 30 Oct 2011 05:06:14 +0000 (05:06 +0000)]
Eliminate vm_phys_bootstrap_alloc(). It was a failed attempt at
eliminating duplicated code in the various pmap implementations.
Micro-optimize vm_phys_free_pages().
Introduce vm_phys_free_contig(). It is fast routine for freeing an
arbitrary number of physically contiguous pages. In particular, it
doesn't require the number of pages to be a power of two.
Use "u_long" instead of "unsigned long".
Bruce Evans (bde@) has convinced me that the "boundary" parameters
to kmem_alloc_contig(), vm_phys_alloc_contig(), and
vm_reserv_reclaim_contig() should be of type "vm_paddr_t" and not
"u_long". Make this change.
marcel [Sun, 30 Oct 2011 02:19:39 +0000 (02:19 +0000)]
Revert rev. 226893: subr_syscall.c is being included from C files and
on amd64 with FREEBSD32 enabled, this means that systrace_probe_func
gets defined twice.
jmg [Sat, 29 Oct 2011 17:28:59 +0000 (17:28 +0000)]
error if /boot/mbr is empty... This can happen on a system like arm
that doesn't have a /boot/mbr, and you touch it to get past the previous
error message...
dougb [Sat, 29 Oct 2011 06:33:12 +0000 (06:33 +0000)]
Fix the text and attribution of the poem "Youth" by Samuel Ullman.
Our copy was a bastardization of the "older" (probably original?) and
shorter version of the poem; and the "newer" version that is better
known, and arguably more popular. Standardize on the latter.
scottl [Sat, 29 Oct 2011 06:26:49 +0000 (06:26 +0000)]
Fix an implicit dependency between the MFI driver and CAM that had grown due
to an API change in CAM. It's once again possible to link a static kernel
with 'mfi' without requiring 'scbus' as well. Ditto for KLD loading.
Submitted by: kib
Reviewed by: ken
MFC after: 3 days
jilles [Fri, 28 Oct 2011 23:02:21 +0000 (23:02 +0000)]
sh: Add test for exit status of for loop without items.
POSIX says the exit status of a for loop without any items shall be 0. There
are no exceptions if the exit status of the previous command was not 0 or if
the item list contains a command substitution with non-zero exit status.