Mateusz Guzik [Tue, 11 Aug 2020 01:34:40 +0000 (01:34 +0000)]
vfs: stricter validation for flags passed to namei in cn_flags
namei de facto expects that the naimeidata object is properly initialized,
but at the same time it mixes consumer-passable and internal flags, while
tolerating this part by explicitly clearing some of them.
Tighten the interface instead.
While here renumber the flags and denote the gap between the 2 variants.
Try to piggy back th renumber on the just bumped __FreeBSD_version.
Rick Macklem [Tue, 11 Aug 2020 00:26:45 +0000 (00:26 +0000)]
Add an argument to newnfs_connect() that indicates use TLS for the connection.
For NFSv4.0, the server creates a server->client TCP connection for callbacks.
If the client mount on the server is using TLS, enable TLS for this callback
TCP connection.
TLS connections from clients will not be supported until the kernel RPC
changes are committed.
Since this changes the internal ABI between the NFS kernel modules that
will require a version bump, delete newnfs_trimtrailing(), which is no
longer used.
Since LCL_TLSCB is not yet set, these changes should not have any semantic
affect at this time.
Mark Johnston [Mon, 10 Aug 2020 20:34:45 +0000 (20:34 +0000)]
Check the UMA zone's full bucket cache before short-circuiting an alloc.
The global "bucketdisable" flag indicates that we are in a low memory
situation and should avoid allocating buckets. However, in the
allocation path we were checking it before the full bucket cache and
bailing even if the cache is non-empty. Defer the check so that we have
a shot at allocating from the cache.
This came up because M_NOWAIT allocations from the buf trie node zone
must always succeed. In one scenario, all of the preallocated trie
nodes were in the bucket list, and a new slab allocation could not
succeed due to a memory shortage. The short-circuiting caused an
allocation failure which triggered a panic.
Reported by: pho
Reviewed by: cem
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D25980
Michael Tuexen [Mon, 10 Aug 2020 20:24:48 +0000 (20:24 +0000)]
Fix the following issues related to the TCP SYN-cache:
* Let the accepted TCP/IPv4 socket inherit the configured TTL and
TOS value.
* Let the accepted TCP/IPv6 socket inherit the configured Hop Limit.
* Use the configured Hop Limit and Traffic Class when sending
IPv6 packets.
- Do not rely on U-Boot for clocks configuration, enable and set frequencies
in the driver's attach method.
- Adjust MAC settings according to detected linespeed on RK3399 and RK3328.
- Add support for RMII PHY mode on RK3328.
Reviewed by: manu
Differential Revision: https://reviews.freebsd.org/D26006
Dimitry Andric [Mon, 10 Aug 2020 16:55:54 +0000 (16:55 +0000)]
Follow-up to r358851 (llvm-project 10.0.0-rc3 import), where I added
subdirectories for compiler-rt's internal fuzzer, profile and xray
headers, but forgot to add installing those headers themselves.
Bjoern A. Zeeb [Mon, 10 Aug 2020 10:58:43 +0000 (10:58 +0000)]
MC: add a note with reference to the discussion and history as-to why we
are where we are now. The main thing is to try to get rid of the delayed
freeing to avoid blocking on the taskq when shutting down vnets.
X-Timeout: if you still see this before 14-RELEASE remove it.
Ryan Moeller [Sun, 9 Aug 2020 16:27:28 +0000 (16:27 +0000)]
Move ifconfig SFP status functionality into libifconfig
libifconfig_sfp.h provides an API in libifconfig for querying SFP module
properties, operational status, and vendor strings, as well as descriptions
of the various fields, string conversions, and other useful helpers for
implementing user interfaces.
SFP module status is obtained by reading registers via an I2C interface.
Descriptions of these registers and the values therein have been collected
in a Lua table which is used to generate all the boilerplace C headers and
source files for accessing these values, their names, and descriptions.
The generated code is fully commented and readable.
This is the first use of libifconfig in ifconfig itself. For now, the
scope remains very limited. Over time, more of ifconfig will be replaced
with libifconfig.
Some minor changes to the formatting of ifconfig output have been made:
- Module memory hex dumps are indented one extra space as a result of using
hexdump(3) instead of a bespoke hex dump function.
- Media descriptions have an added two-character short-name in parenthesis.
- QSFP modules were incorrectly displaying TX bias current as power. Now
TX channels display bias current, and this change has been made for both
SFP and QSFP modules for consistency.
A Lua binding for libifconfig including this functionality is implemented
but has not been included in this commit. The plan is for it to be
committed after dynamic module loading has been enabled in flua.
Michael Tuexen [Sat, 8 Aug 2020 19:39:38 +0000 (19:39 +0000)]
Improve the ECN negotiation when the TCP SYN-cache is used by making
sure that
* ECN is disabled if the client sends an non-ECN-setup SYN segment.
* ECN is disabled is the ECN-setup SYN-ACK segment is retransmitted more
than net.inet.tcp.ecn.maxretries times.
Alex Richardson [Sat, 8 Aug 2020 10:05:27 +0000 (10:05 +0000)]
Fix i386 build of chpass after r363992
My change to allow bootstrapping pwd_mkdb (r363992) resulted in i386 build
failures because the bootstrap header was being included in non-bootstrap chpass.
Dropping the no longer required pwd_mkdb include path from chpass fixes
the build, but to be certain that the failure doesn't get re-introduced,
I've also moved the bootstrap pwd.h into a subdirectory so that adding
-I${SRCTOP}/usr.sbin/pwd_mkdb doesn't pull it in.
Warner Losh [Fri, 7 Aug 2020 16:26:56 +0000 (16:26 +0000)]
The practice of creating symbolic links is somewhat fragile. Always
make copies instead.
There's too many times that we can't run the new binaries with old
libraries. Making the links when things are known to be 'safe' is a
nice optimization, but a copy of all the binaries is only 30MB, so
saving the copies at the cost of increased support when new symbols
are added and used as part of the bootstrap seems to be unwise.
There may be additional optimizations possible here, especially for
!FreeBSD hosts. However, that's beyond the scope of the problem I'm
trying to fix with make failing mid-way through an installworld across
change r363679. This optimization there caused us to run a new binary
with an old library once a new make was installed due to the symbolic
link. One could just copy make, but then other binaries fail as well,
so rather than play whack-a-mole, I opted to take us back to the old
way. Before r340157 or so we did copies (thogh of a lot fewer
artifacts), and we didn't have issues like this.
Implement radix_tree_store() in the LinuxKPI for use with the coming
extensible arrays implementation.
While at it add some more comments explaining the current
radix_tree_insert() function and make sure to clean the root node when
the radix tree reaches the maximum height. This can happen if the
index passed is too big when the tree is empty.
The radix_tree_store() function is basically a copy of the
radix_tree_insert() function with some added functionality.
The radix_tree_store() function is local to FreeBSD and does not yet
exist in Linux.
Alex Richardson [Fri, 7 Aug 2020 16:04:21 +0000 (16:04 +0000)]
Fix linker error in libuutil with recent LLVM
Not marking the function as static can result in a linker error:
undefined reference to __assfail [--no-allow-shlib-undefined]
I noticed this error after updating our CHERI LLVM to the latest upstream
LLVM HEAD revision.
This change effectively reverts r329984 and marks dmu_buf_init_user as
static (which keeps the GCC build happy).
Alex Richardson [Fri, 7 Aug 2020 16:04:06 +0000 (16:04 +0000)]
makefs: Drop unnecessary sys/clock.h include
This breaks the build on macOS where this header doesn't exist. I could
also add a compat header to tools/build/cross-build but since it's not
needed removing it seems like the better solution.
Add full support support for dynamic allocation and freeing of epoch's.
Make sure to reclaim epoch structures when they are freed to support
dynamic allocation and freeing of epoch structures.
While at it, move the 64 supported epoch control structures to the
static memory domain. This overall simplifies the management and
debugging of system epoch's.
Bjoern A. Zeeb [Fri, 7 Aug 2020 15:13:53 +0000 (15:13 +0000)]
IPV6_PKTINFO support for v4-mapped IPv6 sockets
When using v4-mapped IPv6 sockets with IPV6_PKTINFO we do not
respect the given v4-mapped src address on the IPv4 socket.
Implement the needed functionality. This allows single-socket
UDP applications (such as OpenVPN) to work better on FreeBSD.
Bjoern A. Zeeb [Fri, 7 Aug 2020 12:24:23 +0000 (12:24 +0000)]
net80211/ifconfig: print hardware device name for wlan interfaces
Add IEEE80211_IOC_IC_NAME to query the ic_name field and in ifconfig
to print the parent interface again. This functionality was lost
around r287197. It helps in case of multiple wlan interfaces and
multiple underlying hardware devices to keep track which wlan
interface belongs to which physical device.
iflib: netmap: don't increment ifl_cidx on the wrong free list
Netmap only uses free list 0 to keep it consistent with its
one-to-one mapping between each netmap ring and a device RX
(or TX) queue.
However, the current iflib_netmap_rxsync() routine was
mistakenly updating the ifl_cidx field of both free lists.
Alex Richardson [Thu, 6 Aug 2020 20:46:13 +0000 (20:46 +0000)]
Allow bootstrapping pwd_mkdb on Linux/macOS
We need to provide a struct passwd that is compatible with the target
system and this is not the case when cross-building from macOS/Linux.
It should also be a problem when bootstrapping for an i386 target from a
FreeBSD amd64 host since time_t does not match across those systems.
However, pwd_mkdb always truncates integer values to 32-bit so this
difference does not result in different databases.
Alex Richardson [Thu, 6 Aug 2020 20:44:40 +0000 (20:44 +0000)]
ctfmerge: Fix missing pthread_cond_init()
This does not appear to matter on FreeBSD or Linux, but when building an
amd64 kernel on macOS I was seeing infinite loops in ctfmerge.
It turns out the loop in wip_save_work() was looping forever due to
pthread_cond_wait() always returning -EINVAL.
Dimitry Andric [Thu, 6 Aug 2020 20:31:50 +0000 (20:31 +0000)]
Fix clang 11 -Wformat warnings in yp_mkdb:
usr.sbin/yp_mkdb/yp_mkdb.c:91:40: error: format specifies type 'char *' but the argument has type 'void *' [-Werror,-Wformat]
printf("%.*s %.*s\n", (int)key.size, key.data, (int)data.size,
~~~~ ^~~~~~~~
usr.sbin/yp_mkdb/yp_mkdb.c:92:7: error: format specifies type 'char *' but the argument has type 'void *' [-Werror,-Wformat]
data.data);
^~~~~~~~~
Alexander Motin [Thu, 6 Aug 2020 19:16:11 +0000 (19:16 +0000)]
Add CTL support for REPORT IDENTIFYING INFORMATION command.
It allows to report to initiator LU identifying information, preset via
"ident_info" and "text_ident_info" options.
Unfortunately it is impossible to implement SET IDENTIFYING INFORMATION,
since we have no persistent storage it requires, so the information is
read-only for initiator and has to be set out-of-band.
Brandon Bergren [Thu, 6 Aug 2020 17:49:19 +0000 (17:49 +0000)]
[POWERPC] Fix ppc64 makecontext() parameter overflow handling.
On ELFv2, the overflow parameters in the stack frame are at a different offset
from sp than ELFv1. Adjust code to use the correct offset in all cases.
This had resulted in argv[8] and up being copied to the incorrect address
in the new context's initial stack frame.
This is not necessarily the only bug in this function, I need to do a full
review still and ensure the rest of the math is sane for ELFv2 stack frames.
Reported by: pherde (Probably. My notes are a bit unclear.)
Reviewed by: jhibbits (in irc)
Sponsored by: Tag1 Consulting, Inc.
Emmanuel Vadot [Thu, 6 Aug 2020 16:14:43 +0000 (16:14 +0000)]
pkgbase: Add the possibility to choose the output dir
The output dir is set to ${REPODIR}/${PKG_ABI}/${PKG_VERSION} now.
Add the possibility to specify the last componant and set it by default
to ${PKG_VERSION} as before.
This is useful for tests and also for building packages with the same
PKG_VERSION provided to check differences.
Emmanuel Vadot [Thu, 6 Aug 2020 16:13:54 +0000 (16:13 +0000)]
pkgbase: Add a new variable PKG_TIMESTAMP
libarchive uses the SOURCE_DATE_EPOCH env variable to set the date of file
in an archive, this is useful for reproducibility.
Add a variable name PKG_TIMESTAMP that take a epoch time and set SOURCE_DATE_EPOCH
to this. By default it is the current time so no changes here.
Mark Johnston [Thu, 6 Aug 2020 15:43:15 +0000 (15:43 +0000)]
Clean up reassignbuf() and buf_vlist_remove() a bit.
- Convert panic() calls to INVARIANTS-only assertions. The PCTRIE code
provides some of the same protection since it will panic upon an
attempt to remove a non-resident buffer.
- Update the comment above reassignbuf() to reflect reality.
Reviewed by: cem, kib, mjg
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D25965
Mark Johnston [Thu, 6 Aug 2020 15:42:59 +0000 (15:42 +0000)]
Remove the vfs.reassignbufcalls counter and sysctl.
As the 20-year old comment above it suggests, the counter is of dubious
value. Moreover, the (global) counter was not updated precisely and
hurts scalability.
Reviewed by: cem, kib, mjg
MFC after: 1 week
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D25965
The logic is implemented as three sub-drivers.
The parent driver claims the USB device and creates two child devices.
One acts as a GPIO controller and the other is an I2C controller.
Tested with CP2112 revision F02.
Both features seem to work.
HTU21 sensor was used as an I2C slave.
Mateusz Guzik [Wed, 5 Aug 2020 19:15:59 +0000 (19:15 +0000)]
pipe: reduce atime precision
The routine is called on successful write and read, which on pipes happens a
lot and for small sizes.
Precision provided by default seems way bigger than necessary and it causes
problems in vms on amd64 (it rdtscp's which vmexits). getnanotime seems to
provide the level roughly in lines of Linux so we should be good here.
Sample result from will-it-scale pipe1_processes -t 1 (ops/s):
before: 426464
after: 3247421
Note the that atime handling for named pipes is broken with and without the
patch. The filesystem code is never used for updating atime and never looks
at the updated field. Consequently, while there are no provisions added to
handle named pipes separately, the change is a nop for that case.
Ruslan Bukin [Wed, 5 Aug 2020 19:11:31 +0000 (19:11 +0000)]
o Add machine/iommu.h and include MD iommu headers from it,
so we don't ifdef for every arch in busdma_iommu.c;
o No need to include specialreg.h for x86, remove it.