Alexander Motin [Thu, 23 May 2024 16:20:37 +0000 (12:20 -0400)]
Fix scn_queue races on very old pools
Code for pools before version 11 uses dmu_objset_find_dp() to scan
for children datasets/clones. It calls enqueue_clones_cb() and
enqueue_cb() callbacks in parallel from multiple taskq threads.
It ends up bad for scan_ds_queue_insert(), corrupting scn_queue
AVL-tree. Fix it by introducing a mutex to protect those two
scan_ds_queue_insert() calls. All other calls are done from the
sync thread and so serialized.
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Brian Atkinson <batkinson@lanl.gov>
Signed-off-by: Alexander Motin <mav@FreeBSD.org>
Sponsored by: iXsystems, Inc.
Closes #16162
PR: 278414
Brooks Davis [Thu, 23 May 2024 16:14:36 +0000 (17:14 +0100)]
pcm: centralize 32-bit ioctl compat
Move all handlng of struct sndstioc_nv_arg(32) to sndstat_ioctl() and
make the functions that actually do the work take a buffer and size or
size pointer. The 32-bit compat work is minimal so just inline it.
Remove checks that we've got a 32-bit process for 32-bit ioctls. We
don't check that default ioctls are from 64-bit processes on 64-bit
systems.
Stop treating size 0 as unknown size in vnode_create_vobject().
Whenever file is created, the vnode_create_vobject() function will
try to determine its size by calling vn_getsize_locked() as size 0
is ambigious: it means either the file size is 0 or the file size
is unknown.
Introduce special value for the size argument: VNODE_NO_SIZE.
Only when it is given, the vnode_create_vobject() will try to obtain
file's size on its own.
Introduce dedicated vnode_disk_create_vobject() for use by
g_vfs_open(), so we don't have to call vn_isdisk() in the common case
(for regular files).
Zhenlei Huang [Thu, 23 May 2024 04:10:17 +0000 (12:10 +0800)]
freebsd-update: Correctly check if pkg(8) is present
On systems without pkg(8) installed, `command -v pkg` will return
success and falsely report that pkg(8) is present. Fix that by checking
via the `pkg -N` form.
This is missing from the final revision of D39695.
Alan Cox [Mon, 13 May 2024 06:39:28 +0000 (01:39 -0500)]
arm64: set ATTR_CONTIGUOUS on the DMAP's L2 blocks
On systems configured with 16KB pages, this change creates 1GB page
mappings in the direct map where possible. Previously, the largest page
size that was used to implement the direct map was 32MB. Similarly, on
systems configured with 4KB pages, this change creates 32MB page
mappings, instead of 2MB, in the direct map where 1GB is too large.
Implement demotion on L2C (32MB/1GB) page mappings within the DMAP.
Update sysctl vm.pmap.kernel_maps to report on L2C page mappings.
sound: Fix minchn, maxchn and fmts in sndstat_get_caps()
The current implementation (incorrectly) passes the channel encoding
value to AFMT_CHANNEL(), which will always return 0, since the channel
number bits are masked out by AFMT_ENCODING().
Also add missing fmts initialization and aggregate encoding formats into
it directly.
Sponsored by: The FreeBSD Foundation
MFC after: 1 day
Reviewed by: dev_submerge.ch
Differential Revision: https://reviews.freebsd.org/D45312
mixer(8): Ignore mixer_open() failures for the -a option
The most likely reason mixer_open() will fail is because either the
device doesn't exist, or because it is disabled, so there is not reason
to kill the application. Instead, continue and print the rest of the
enabled mixers.
PR: 277615
Sponsored by: The FreeBSD Foundation
MFC after: 1 day
Reviewed by: dev_submerge.ch
Differential Revision: https://reviews.freebsd.org/D45151
sound: Handle unavailable devices in various OSS IOCTLs
mixer(8)'s -a option is used to print information about all mixer
devices in the system. To do this, it loops from 0 to
mixer_get_nmixers(), and tries to open "/dev/mixer%d". However, this
approach doesn't work when there are disabled/unregistered mixers in the
system, or when an audio device simply doesn't have a mixer.
mixer_get_nmixers() calls SNDCTL_SYSINFO and returns
oss_sysinfo->nummixers, whose value is the number of currently _enabled_
mixers only. Taking the bug report mentioned below (277615) as an
example, suppose a system with 8 mixer devices total, but 3 of them are
either disabled or non-existent, which means they will not show up under
/dev, meaning we have 5 enabled mixer devices, which is also what the
value of oss_sysinfo->nummixers will be. What mixer(8) will do is loop
from 0 to 5 (instead of 8), and start calling mixer_open() on
/dev/mixer0, up to /dev/mixer4, and as is expected, the first call will
fail right away, hence the error shown in the bug report.
To fix this, modify oss_sysinfo->nummixers to hold the value of the
maximum unit in the system, which, although not necessarily "correct",
is more intuitive for applications that will want to use this value to
loop through all mixer devices.
Additionally, notify applications that a device is
unavailable/unregistered instead of skipping it. The current
implementations of SNDCTL_AUDIOINFO, SNDCTL_MIXERINFO and
SNDCTL_CARDINFO break applications that expect to get information about
a device that is skipped. Related discussion can be found here:
https://reviews.freebsd.org/D45135#1029526
It has to be noted, that other applications, apart from mixer(8), suffer
from this.
PR: 277615
Sponsored by: The FreeBSD Foundation
MFC after: 1 day
Reviewed by: dev_submerge.ch
Differential Revision: https://reviews.freebsd.org/D45256
sound: Separate implementations for SNDCTL_AUDIOINFO[_EX] and SNDCTL_ENGINEINFO
FreeBSD's implementation of SNDCTL_AUDIOINFO[_EX] and SNDCTL_ENGINEINFO
does not exactly work as intended. The problem is essentially that both
IOCTLs return the same information, while in fact the information
returned currently by dsp_oss_audioinfo() is what _only_
SNDCTL_ENGINEINFO is meant to return.
This behavior is also noted in the OSS manual [1] (see bold paragraph in
"Audio engines and device files" section), but since e8c0d15a64fa
("sound: Get rid of snd_clone and use DEVFS_CDEVPRIV(9)") we can
actually fix this, because we now expose only a single device for each
soundcard, and create the engines (channels) internally.
SNDCTL_ENGINEINFO will now report info about all channels in a given
device, and SNDCTL_AUDIOINFO[_EX] will only report information about
/dev/dspX.
To make this work, we also have to modify the SNDCTL_SYSINFO IOCTL to
report the number of audio devices and audio engines correctly.
While here, modernize the minimum and maximum channel counting in both
SNDCTL_AUDIOINFO[_EX] and SNDCTL_ENGINEINFO. Currently these IOCTLs will
report only up to 2 channels, which is no longer the case.
Bjoern A. Zeeb [Mon, 13 May 2024 21:03:31 +0000 (21:03 +0000)]
LinuxKPI: 802.11: rename chanctx_conf for consistency
We used to call the struct ieee80211_chanctx_conf variable "conf"
in some places but that becomes confusing with bss_conf and other
"conf" bits. On the vif it is already called chanctx_conf thus also
rename it on the internal struct lkpi_chanctx and for our variables
in the implementation.
This should not have any external visibility.
No functional changes intended.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D45185
Bjoern A. Zeeb [Wed, 22 May 2024 02:24:51 +0000 (02:24 +0000)]
LinuxKPI: 802.11: change teardown order to avoid iwlwifi firmware crashes
While the previous order worked well for iwlwifi 22000 and later chipsets
(AXxxx, BE200), earlier chipsets had trouble and ran into firmware crashes.
Change the teardown order to avoid these problems. The inline comments
in lkpi_sta_run_to_init() (and lkpi_disassoc()) try to document the new
order and also the old problems we were seeing (too early sta removal or
silent non-removal) leading to follow-up problems.
There is a possible further problem still lingering but a lot harder to
trigger (see comment in review) and likely related to some other doings
so we'll track it separately.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
PR: 275255
Tested with: AX210, 8265 (bz); 9260 (Bakul Shah)
Differential Revision: https://reviews.freebsd.org/D45293
Add function to OSD to get values without taking the lock.
There are some cases of OSD use where the value is only initialized once
at a point where successive access of the value can be done so safely
without the need to take the lock.
Michael Tuexen [Wed, 22 May 2024 18:36:24 +0000 (20:36 +0200)]
blackhole.4: improve man page
Not the connection is dropped, but the incoming SYN segment.
Reviewed by: concussious.bugzilla_runbox.com
MFC after: 3 days
Sponsored by: Netflix, Inc.
Differential Revision: https://reviews.freebsd.org/D45296
Andrew Turner [Wed, 22 May 2024 15:54:32 +0000 (15:54 +0000)]
arm64: Allow userspace to be built with PAC and BTI
Add the WITH/WITHOUT_BRANCH_PROTECTION build flags. This can be used
to enable the use of pointer authentication (FEAT_PAuth) and branch
target identification (FEAT_BTI) in userspace.
The kernel already handles both of these is userspace, we just need
to enable it.
Leave disabled for a short period for this to settle before enabling.
Reviewed by: emaste
Sponsored by: Arm Ltd
Differential Revision: https://reviews.freebsd.org/D42596
Ruslan Bukin [Wed, 22 May 2024 14:51:45 +0000 (15:51 +0100)]
riscv: Fix SSTC extension support
From the SSTC spec:
"If the stimecmp (supervisor-mode timer compare) register is implemented,
then STIP is read-only in mip and reflects the supervisor-level timer
interrupt signal resulting from stimecmp. This timer interrupt signal
is cleared by writing stimecmp with a value greater than the current time
value."
This fixes operation in Spike with sstc extension enabled.
Example:
spike --isa RV64IMAFDCH_zicntr_zihpm_sstc
Piotr Kubaj [Tue, 21 May 2024 10:28:35 +0000 (12:28 +0200)]
usr.bin/top: fix displaying load average for loads of at least 100
After top registers load average of at least 100 which then gets reduced to
below 100, there are left stray digits.
Supporting load over 100 requires increasing the width only to 6, but since
we support over 1000 CPU's now, let's increase it to 7.
rc.d: add a service jails config to all base system services
This gives more permissions to services (e.g. network access to
services which require this) when they are started as an automatic
service jail.
The sshd patch is important for the sshd-related functionality as
described in the man-page in the service jails part.
The location of the added env vars is supposed to allow overriding them
in rc.conf, and to hard-disable the use of svcj for some parts where it
doesn't make sense or will not work.
Only a subset of all of the services are fully tested (I'm running this
since more than a year with various services started as service jails).
The untested parts should be most of the time ok, in some edge-cases
more permissions are needed inside the service jail.
Differential Revision: https://reviews.freebsd.org/D40371
This takes a rc.d-service and starts it in a jail which shares the same
root-path as the host (or parent jail) and may inherit the network from
the host (or parent jail). Per service there is the possibility to
specify some arguments which give more permissions (e.g. netv4, netv6,
sysvipc...).
Reviewed by: bcr (man page)
Relnotes: yes
Differential Revision: https://reviews.freebsd.org/D40370
mac_do: add a new MAC/do policy and mdo(1) utility
This policy enables a user to become another user without having to be
root (hence no setuid binary). it is configured via rules using sysctl
security.mac.do.rules
For example:
security.mac.do.rules=uid=1001:80,gid=0:any
The above rule means the user identifier by the uid 1001 is able to
become user 80
Any user of the group 0 are allowed to become any user on the system.
The mdo(1) utility expects the MAC/do policy to be installed and its
rules defined.
Reviewed by: des
Differential Revision: https://reviews.freebsd.org/D45145
Dmitry Salychev [Tue, 21 May 2024 20:25:59 +0000 (22:25 +0200)]
arm64: Fixed IOMMU compilation errors
These are missing changes after 1228b93b410a when ref_count was
removed from bus_dma_tag_common and 1e3f42b6bad5, when the address
arguments were switched to pointers.
Reviewed by: jhb
MFC after: 3 days
Differential Revision: https://reviews.freebsd.org/D45289
Andrew Turner [Wed, 22 May 2024 08:19:38 +0000 (08:19 +0000)]
pci: Fix pci_host_generic_acpi with gcc
In pci_host_generic_acpi.c we loop over pci_acpi_quirks to check if
we need to handle any quirks. GCC doesn't like the terminatin as it
sets a fixed width string to 0.
As this the array is only ever used in this file change to use nitems
to find when to stop the loop.
Andrew Turner [Wed, 22 May 2024 08:19:06 +0000 (08:19 +0000)]
arm64: Use the pointer auth register defines
When building with gcc it complains the pointer authentication
registers aren't valid with the architecture level we are targeting.
Fix this by using the alternative spelling of these registers accesses
through MRS_REG_ALT_NAME.
Reviewed by: jhb
Sponsored by: Arm Ltd
Differential Revision: https://reviews.freebsd.org/D45263
Andrew Turner [Wed, 22 May 2024 08:18:39 +0000 (08:18 +0000)]
arm64: Use the UL macro in TCR_EL1 defines
While clang can handle numbers with a UL suffix in assembly files
gcc/gas is unable to. Switch to use the UL macro for TCR_EL1 defines as
some are used in locore.S
Andrew Turner [Wed, 22 May 2024 08:18:26 +0000 (08:18 +0000)]
sys/sys: Fix __builtin_is_aligned fallback
When the compiler doesn't provide __builtin_is_aligned we use
macro as a fallback. The macro was missing brackets around one
argument. This could lead to incorrect results when the argument is
more complex than a single stagement.
Andrew Turner [Wed, 22 May 2024 08:18:10 +0000 (08:18 +0000)]
csu: Find the main pointer through the GOT
Use the Global Offset Table to find the location of main in crt1. With
lld the old code would point to main@plt, however ld.bfd fails to link
when main is in a shared library.
Fix this by using the GOT address to find main as it works with both
lld and bfd.
Reviewed by: jrtc27
Sponsored by: Arm Ltd
Differential Revision: https://reviews.freebsd.org/D45259
Kristof Provost [Tue, 21 May 2024 11:31:13 +0000 (13:31 +0200)]
if_vlan: handle VID conflicts
If we fail to change the vlan id we have to undo the removal (and vlan id
change) in the error path. Otherwise we'll have removed the vlan object from the
hash table, and have the wrong vlan id as well. Subsequent modification attempts
will then try to remove an entry which doesn't exist, and panic.
Undo the vlan id modification if the insertion in the hash table fails, and
re-insert it under the original vlan id.
Elliott Mitchell [Fri, 10 May 2024 16:13:39 +0000 (09:13 -0700)]
kern/rman: mark rman get functions as taking constants
The arguments are left completely unchanged by these functions. This
allows passing constant pointers for verifying ownership, but not
modifying the contents.
Not once has rman_reserve_resource_bound() ever been used. There are
though several uses of RF_ALIGNMENT. In light of this remove this
extra and leave the actually used portion in place.
Rather than hard-code the function name, use __func__ instead. Apply
some style and adjust indentation as appropriate. Remove the no longer
required braces.
kern/rman: update DPRINTF() macro, avoid semicolon swallowing match function
Using a variadic macro allows passing everything properly to printf().
Using the do { } while(0) construct ensures the macro acts like any
other single statement. This shows just how long some of this has
existed.
Pierre Pronchery [Thu, 16 May 2024 14:34:34 +0000 (16:34 +0200)]
libnvmf: avoid resource leak
In nvmf_host_fetch_discovery_log_page(), the log variable may have been
allocated on the heap during the first loop cycle, and should be
free()'d before exiting upon errors.
Reported by: Coverity
CID: 1545034
Sponsored by: The FreeBSD Foundation
Bjoern A. Zeeb [Mon, 13 May 2024 17:40:26 +0000 (17:40 +0000)]
LinuxKPI: add FIELD_PREP_CONST()
Add FIELD_PREP_CONST() like FIELD_PREP() without any extra checks likely
expected on this version in Linux. This is called by an updated wireless
driver.
Sposnored by: The FreeBSD Foundation
MFC after: 3 days
Reviewed by: emaste
Differential Revision: https://reviews.freebsd.org/D45180
Allan Jude [Tue, 21 May 2024 22:40:12 +0000 (22:40 +0000)]
ftpd: stop using -g flag for /bin/ls
In 3bfbb521 the behaviour of ls was changed such that -g was no longer
a noop for compatibility with BSD 4.3, but instead changed the output
of long mode to exclude the owner of the file and display only the
group.
Update how FTPd invokes ls to restore the previous behaviour
Reported-by: Andrew Fengler <andrew.fengler@scaleengine.com>
Reviewed-by: jrtc27, des, imp
MFC after: 3 days
Sponsored-by: ScaleEngine Inc.
Fixes: 3bfbb521fef5 ("ls: Improve POSIX compatibility for -g and -n.")
Bjoern A. Zeeb [Mon, 13 May 2024 17:43:25 +0000 (17:43 +0000)]
LinuxKPI: add kvmemdup()
Add kvmemdup() as a variant of kmemdup(). While currently it could
just call kmemdup() we duplicate the code and use kvmalloc() in case
someone will change the implementation of kvmalloc/kvfree in slab.h.
This is used by an updated wireless driver.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Reviewed by: emaste
Differential Revision: https://reviews.freebsd.org/D45181
Bjoern A. Zeeb [Tue, 21 May 2024 21:58:47 +0000 (21:58 +0000)]
LinuxKPI: 802.11: fix for_each_sta_active_link()
Likely a c&p error from for_each_vif_active_link() to
for_each_sta_active_link().
We are checking the nitems on the vif instead of the sta in this macro.
Function wise there is no difference as the arrays are the same size
but for correctness fix this.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
Chris Moerz [Tue, 21 May 2024 18:10:11 +0000 (15:10 -0300)]
glabel.8: Describe cases related to permissions / existing mounts
Specially, note some requirements for label changes:
- glabel requires write permission on device
- filesystems first need to be unmounted for new labels to persist
across reboots
- if the affected device node holds the filesystem root, single-user
mode with r/o mount will be required.
Also, while here, apply some formatting tweaks.
PR: 276724
Reported by: Alex Matei <matei35@yahoo.com>
Reviewed by: gbe, jrm, Alexander Ziaee <concussious@runbox.com>
Differential Revision: https://reviews.freebsd.org/D44394
Ryan Libby [Tue, 21 May 2024 17:21:50 +0000 (10:21 -0700)]
getblk: fail faster with GB_LOCK_NOWAIT
If we asked not to wait on a lock, and then we failed to get a buf lock
because we would have had to wait, then just return the error. This
avoids taking the bufobj lock and a second trip to lockmgr.
John Baldwin [Tue, 21 May 2024 16:42:12 +0000 (09:42 -0700)]
cryptocheck: Don't test Chacha20-Poly1305 with an IV size of 8
OpenSSL 3.0+ doesn't support an IV size of 8 either for the Chacha20
stream cipher or the AEAD combination with Poly1305. This did work
previously with OpenSSL 1.1.
Zhenlei Huang [Tue, 21 May 2024 16:35:01 +0000 (00:35 +0800)]
ifconfig: Redo fix vlan/vlanproto reconfiguration
When the if_vlan(4) interface has not been fully configured, i.e., a
bare interface without a physical interface associated with it,
retrieving the current settings of it and unconditionally overwriting
`params` will result in losing vlandev settings in `params`. That will
lead to failing to associate the if_vlan(4) interface with the requested
physical interface and the false report 'both vlan and vlandev must be
specified'.
Fix that by checking if the vlan interface has been fully configured.
The basic VLAN test is slightly modified to cover this case.