sys_attrs_generic(5)sys_attrs_generic(5)NAMEsys_attrs_generic - system attributes for the generic kernel subsystem
DESCRIPTION
This reference page describes the attributes for the Generic (generic)
kernel subsystem. See sys_attrs(5) for general information about set‐
ting system attributes.
In the following list, attributes whose names are preceded by an aster‐
isk (*) can be modified at run time as well as at boot time.
A value that is set and used internally to support a rolling
upgrade operation in a TruCluster environment. Do not modify
manually.
A value that is set and used internally to support a rolling
upgrade operation in a TruCluster environment. Do not modify
manually.
The kernel buffer size in bytes.
Default value: If the binlog_buffer_size is not explicitly spec‐
ified, the value for kernel buffer size is assigned at boot time
and is based on the amount of physical memory installed on the
system. The value assigned automatically ranges from 32 kilo‐
bytes to 1 megabyte.
Minimum value: 8192 (bytes, or 8 KB)
Maximum value: 1,048,576 (bytes, or 1 MB)
The arguments that are passed from osf_boot to the kernel. This
attribute value is query only.
The name of the kernel that was loaded by osf_boot. This
attribute value is query only.
A value that indicates whether the system is (1) or is not (0)
an actively running member system in a TruCluster environment.
This query-only value is always set to 0 for a standalone sys‐
tem.
A value that indicates whether cluster kernel components are (1)
or are not (0) configured in the kernel. This query-only value
is always set to 0 for a standalone system.
A value that enables (1) or disables (0) a compressed dump file.
The type of zero compression implemented for compression saves
substantial disk space but still allows the dump to be read by
dbx and other operating system tools. See expected_dump_com‐
pression for information about setting the compression level.
Default value: 1 (enabled)
You might want to change the default value if you consistently
use tools, such as those from a third party, that do not support
compressed dump files. See the System Administration and the
Kernel Debugging manuals for more information about generating
and working with dump files.
See the dump_savecnt entry for overview information about dump-
related attributes.
A value for the bit mask that determines which CPUs are started
in a multiprocessor system. The lowest order bit (bit 0) in
this mask corresponds to CPU0. The next highest order bit corre‐
sponds to CPU1, and so on, up to bit 63. The bits in these posi‐
tions can be set to 0 (disabled) or 1 (enabled) for the corre‐
sponding CPU slots. The decimal or hexadecimal value specified
for cpu_enable_mask corresponds to the binary value that has 0
and 1 in the correct ordinal positions for the CPUs you want to
enable or disable.
Default value: The hex value corresponding to the slots of CPUs
that are actually enabled. On single-CPU systems with the CPU in
slot 0, the default value is 0x1.
Minimum value: 0
Maximum value: 0xffffffffffffffff (or -1)
Specifying -1 for this attribute means that all CPUs present in
the system are allowed to be started by the master CPU. Specify‐
ing 0 for this attribute enables uniprocessor operation in a
multi-CPU system. Other values enable and disable specific CPUs,
depending on what the system supports.
The following table illustrates the relationships between bit
settings and cpu_enable_mask values for enabling a single CPU.
It is generally easier to determine and enter hexadecimal val‐
ues, so only a few significant decimal equivalents are shown:
Bit Dec. Hex Single
63... 8 7 6 5 4 3 2 1 0 Value Value CPU Enabled
______________________________________________________________________
0 ... 0 0 0 0 0 0 0 0 0 0 0 Special
meaning * 0 ... 0 0 0 0 0 0 0 0 1 1 0x1
CPU0 0 ... 0 0 0 0 0 0 0 1 0 0x2 CPU1 0
... 0 0 0 0 0 0 1 0 0 0x4 CPU2 0 ... 0
0 0 0 0 1 0 0 0 0x8 CPU3 0 ... 0 0 0 0
1 0 0 0 0 0x10 CPU4 0 ... 0 0 0 1 0 0 0
0 0 0x20 CPU5 0 ... 0 0 1 0 0 0 0 0 0
0x40 CPU6 0 ... 0 1 0 0 0 0 0 0 0 0x80
CPU7 0 ... 1 0 0 0 0 0 0 0 0 0x100 CPU8
.
.
. 1 ... 0 0 0 0 0 0 0 0 0 0x8000000000000000 CPU63
* The minimum value (0) indicates that only the master (boot)
CPU be enabled. This value is useful if you want to test appli‐
cations in a uniprocessor environment and your test system nor‐
mally runs in multiprocessor mode.
Note
Some systems require that the master CPU be CPU0. For these sys‐
tems, 0, 1, and 0x1 are equivalent entries and are the only val‐
ues you should specify for uniprocessor operation. On these sys‐
tems, 0x1 is always the value set for cpu_enable_mask because
the kernel will overwrite values specified for cpu_enable_mask
that are incompatible with having CPU0 as the master CPU. For
other systems, users can choose the slot position of the master
CPU. Therefore, the result of entering 0 for this attribute can
vary. For example, the actual setting would be 0x2 if CPU1 were
the master CPU and 0x10 if CPU4 were the master CPU.
The following table illustrates the relationships between bit
settings and cpu_enable_mask values for enabling sample combina‐
tions of CPUs:
Bit Dec. Hex Multiple
63... 8 7 6 5 4 3 2 1 0 Value Value CPUs Enabled
_________________________________________________________________________
0 ... 0 0 0 0 0 0 0 1 1 0x3 CPU0, CPU1 0
... 0 0 0 0 0 0 1 0 1 0x5 CPU0, CPU2 0
... 0 0 0 0 0 0 1 1 0 0x6 CPU1, CPU2 0
... 0 0 0 0 0 0 1 1 1 0x7 CPU0, CPU1,
CPU2
. 0 ... 0 0 0 0 0 1 1 1 1 0xf CPU0
to CPU3
. 0 ... 0 0 0 0 1 1 1 1 1 0x1f CPU0
to CPU4
. 0 ... 0 0 0 1 1 1 1 1 1 0x3f CPU0
to CPU5
. 0 ... 0 0 1 1 1 1 1 1 1 0x7f CPU0
to CPU6
. 0 ... 0 1 1 1 1 1 1 1 1 0xff CPU0
to CPU7
. 0 ... 1 1 1 1 1 1 1 1 1 0x1ff CPU0
to CPU8
.
.
. 1 ... 1 1 1 1 1 1 1 1 1 -1 0xffffffffffffffff All
available CPUs *
* The maximum value specifies 1 in all bit positions, thereby
enabling all available CPUs. This value corresponds to
18446744073709551615 or -1 in decimal notation and
0xffffffffffffffff in hexadecimal notation.
You can enable and disable specific combinations of CPUs by
entering values that set and clear appropriate bit positions in
the bit mask. If you suspect that one of the CPUs on a multipro‐
cessor system may be the cause of a system problem, you can try
running the system with different combinations of available CPUs
disabled to help confirm your suspicion. Remember that if your
system is one of those that require CPU0 to be the master CPU,
the kernel does not apply values that map to 0 in bit position
0. In this case, you have to switch another CPU into slot 0
before you can do a test run that disables the CPU originally in
slot 0.
Certain lockmode attribute values override incompatible
cpu_enable_mask settings. If the value of the lockmode
attribute is 0 or 1, all bits in the cpu_enable_mask bit mask
that do not correspond to the master CPU are set to 0. For the
AlphaServer ES47, ES80, and GS1280 systems, the cpus_in_rad
attribute can be set to more than two CPUs to allow the pooling
together of the CPU, memory, and I/O resources of a set of CPUs
and treating it as a single resource affinity domain (RAD).
Like the vm_overflow attribute, this can allow large memory
applications to have access to more memory that is managed as if
it were all physically “local.”
With the default value of cpus_in_rad (zero), every CPU is in
its own RAD. This is equivalent to setting the value to 1. When
the value of cpus_in_rad is set larger than 1, certain configu‐
ration restrictions must be considered: Values for the
cpus_in_rad attribute must be a power of 2. Take caution set‐
ting cpus_in_rad to 64 on a 64 processor system. When the number
of RADs is decreased by increasing the value of cpus_in_rad, the
number of per-RAD locks needed to manage resources also
decreases. This may result in increased lock contention, thereby
causing poor performance or system panics.
The system automatically adjusts the generic locktimeout
attribute if cpus_in_rad is set to 64 on a 64 processor system,
but depending on system load it may need to be manually
increased to avoid locktimeout panics.
The maximum value for locktimeout is 60 seconds. If the value
needs to be increased, do so in 5 second increments. If the max‐
imum value is reached and the system is unstable, reduce the
value of cpus_in_rad and escalate the problem through your sup‐
port channels. “Missing” CPUs are included in the count of
cpus_in_rad.
Consider a system configured with CPUs 0,1,4,5,8,9,12,13. Set‐
ting cpus_in_rad to 2 on this system would result in the follow‐
ing resource affinity domain configuration:
RAD[0] - cpus 0, 1
RAD[2] - cpus 4, 5
RAD[4] - cpus 8, 9
RAD[6] - cpus 12, 13
Setting cpus_in_rad to 4 on this system would result in the fol‐
lowing resource affinity domain configuration:
RAD[0] - cpus 0, 1 (2 and 3 are missing)
RAD[1] - cpus 4, 5 (6 and 7 are missing)
RAD[2] - cpus 8, 9 (10 and 11 are missing)
RAD[3] - cpus 12, 13
Interaction of cpus_in_rad and the rad_gh_regions attributes:
When cpus_in_rad is increased, the number or RADs configured
decreases. If the system is configured with settings for
rad_gh_regions, those settings must also be changed.
Consider a system configured with CPUs 0,1,4,5,8,9,12,13 and
rad_gh_regions configured to allocate 4 GB of granularity hint
memory. With the default setting of cpus_in_rad or cpus_in_rad
set to 1, rad_gh_regions would have the following settings:
rad_gh_regions[0] = 512
rad_gh_regions[1] = 512
rad_gh_regions[4] = 512
rad_gh_regions[5] = 512
rad_gh_regions[8] = 512
rad_gh_regions[9] = 512
rad_gh_regions[12] = 512
rad_gh_regions[13] = 512
If the system is configured to place two CPUs in a RAD
(cpu_in_rad=2), the rad_gh_regions settings would need to be
changed to the following:
rad_gh_regions[0] = 1024
rad_gh_regions[2] = 1024
rad_gh_regions[4] = 1024
rad_gh_regions[6] = 1024
Because of how missing CPUs are handled, if cpus_in_rad is set
to 4, the RADs would still contain only two CPUs (two existing,
two missing) but the RAD numbers change, so rad_gh_regions would
have the following settings:
rad_gh_regions[0] = 1024
rad_gh_regions[1] = 1024
rad_gh_regions[2] = 1024
rad_gh_regions[3] = 1024
Identifies the starting address (virtual or physical) for a
region of exempt memory to be used for writing primary system
core dumps. The specified address must be page aligned.
Default value: 0 (disables writing the dump to an exempt memory
region)
This attribute can be set at run time. The setting of
dump_exmem_addr has no effect unless the dump_exmem_size
attribute is also set to specify the size of the exempt memory
region to contain the dump. If you decide to write system core
dumps to a region of exempt memory, make sure you keep a record
of any run-time settings for dump_exmem_addr and dump_exmem_size
so you will be able to find a crash dump after recovery from a
system failure.
The dump_exmem_addr and dump_exmem_size attributes have no
effect if the setting of the dump_to_memory attribute disables
writing of dumps to memory or if the setting of the dump_savecnt
attribute disables writing of dumps altogether.
See the dump_savecnt entry for overview information about dump-
related attributes.
Determines whether exempt memory pages are included (1) or not
included (0) in the system core dump.
Default value: 0 (exempt memory pages are not included in the
dump)
This attribute can be set at run time.
See the dump_savecnt entry for overview information about dump-
related attributes.
Specifies the size (in bytes) of the exempt memory region to
which system core dumps are written. This value must be a multi‐
ple of the machine word size.
Default value: 0 (disables writing the dump to an exempt memory
region)
This attribute can be set at run time. The setting of this
attribute has no effect unless the dump_exmem_addr attribute is
also set to specify the starting address for the region being
sized. The description of dump_exmem_addr discusses additional
attribute dependencies.
See the dump_savecnt entry for overview information about dump-
related attributes.
Enables or disables the inclusion of kernel text pages. This
attribute only applies when partial dumps are enabled (see par‐
tial_dump and dump_savecnt). Dumped kernels will be larger.
Default value: 0 (disabled)
Minimum value: 0 (disabled)
Maximum value: 1 (enabled)
This value can be changed at run time.
See the dump_savecnt entry for overview information about dump-
related attributes.
A value that limits the number of successful crash dumps that
are generated for a single crash/reboot sequence. Valid values
are: Never generate a crash dump Generate a primary crash dump
(the default) Generate a secondary crash dump
A secondary fault can occur on system failure, in which case, a
secondary crash dump is generated in addition to the primary
crash dump. The secondary crash dump is always written to non-
exempt memory. See the entry for the dump_to_memory attribute
for information that also has impact on whether a secondary dump
is produced and whether it overwrites the primary dump.
System core dumps provide critical troubleshooting information
to your support representative. Although the default value of
the dump_savecnt attribute can be changed at run time, do so
only under direction of your support representative.
See the entries for other dump_* attributes and the com‐
pressed_dump, expected_dump_compression, partial_dump, and
live_dump_* attributes for information about controlling where
system dumps are written, whether dumps are compressed (and at
which level), and the amount and type of information that dumps
include. In general, changes to any of these attributes are
best done under the guidance of your support representative.
See the Kernel Debugging and the System Administration manuals
for more information about creating and working with dump files.
A threshold size that determines whether single-partition or
multiple-partition dumps are created. If a dump will fit on the
primary swap partition and leave space that is equal to this
threshold value, the dump is created as a single-volume dump on
the primary swap partition, even if secondary swap partitions
are available. (See the System Administration and the Kernel
Debugging manuals for details.)
Default value: 16,384
Minimum value: 0
Maximum value: 2,147,483,647
This value can be changed at run time.
See the dump_savecnt entry for overview information about dump-
related attributes.
A value that controls whether primary system core dumps are
written to memory or disk. This attribute can have the follow‐
ing values: Dumps are written only to disk; writing dumps to
memory is disabled. This value also disables writing a secondary
dump (when dump_savecnt is equal to 2). Dumps are written to
disk except in the event of disk failure, in which case they are
written to memory. This is the default behavior. Dumps are
written only to memory when sufficient memory is available; oth‐
erwise, a disk dump might be generated. See the following dis‐
cussion if primary and secondary dumps are both enabled (when
dump_savecnt is equal to 2).
The dump_to_memory value can be changed at run time; however, do
so only under direction of your support representative.
By default, a primary system crash dump that is written to mem‐
ory (dump_to_memory is set to 1) occupies non-exempt memory. If
the dump_savecnt attribute is set to 2 (to enable a secondary
crash dump), the secondary dump is also written to non-exempt
memory. Therefore, setting dump_savecnt to 2 and dump_to_memory
to 1 will cause the secondary dump to clobber the primary dump.
If you want to enable both a primary and a secondary crash dump
and have both dumps available in memory, you can work around
this problem by writing the primary dump to exempt memory. See
the dump_exmem_addr and dump_exmem_size attributes for informa‐
tion about writing primary memory dumps to exempt memory rather
than non-exempt memory.
See the dump_savecnt attribute for information about disabling
the writing of system core dumps altogether and for cross-refer‐
ences to other dump-related attributes.
A value that enables (1) or disables (0) the inclusion of user
page table entries in a crash dump. Normally, user page table
entries contribute no information about the cause of a system
crash.
Default value: 0 (disabled)
This value can be changed at run time. However, you would nor‐
mally change it only when directed to do so by a support repre‐
sentative who is gathering information about a particular prob‐
lem.
A value that enables (1) or disables (0) asynchronous printing
of messages to the console. Asynchronous printing to the console
prevents large numbers of messages from being written to the
console at the same time.
Default value: 1 (enabled)
This value can be changed at run time. Device driver developers
need to set enable_async_printf to 0 when debugging driver prob‐
lems because they need to see the results of kernel printf()
calls at the time of execution.
See the dump_savecnt entry for overview information about dump-
related attributes.
The level of compression that a dump is typically expected to
achieve.
Default value: 500
Minimum value: 0
Maximum value: 1000
This value can be changed at run time. See the System Adminis‐
tration manual for instructions on how to calculate the typical
dump compression level for your system.
See the dump_savecnt entry for overview information about dump-
related attributes.
A value that allows programmers to override file permissions
when UNIX domain sockets are created by a bind() call.
When insecure_bind is set to a value other than 0, UNIX domain
sockets are always created with mode 0777 (the umask is
ignored). This prevents bind: permission denied errors when the
bind() call executes.
Default value: 0, meaning that file permission (umask) is not
ignored
This value can be changed at run time. See bind(2) for more
information about creating domain sockets.
In a TruCluster environment, the value of this attribute must be
the same on all member systems.
See kmem_debug.
A value that enables (1) or disables (0) behavior useful for
debugging problems in kernel memory allocation.
Default value: 0
If enabled (1), each time the kernel memory allocator allocates
or deallocates memory in the kernel memory pool, the system
checks whether the operation is performed correctly. If the ker‐
nel memory pool is in a corrupt state, the system crashes and
provides useful debugging information. This attribute, along
with the kmem_audit_count, kmem_debug_size_mask, kmem_pro‐
tected_kmempercent, kmem_protected_lowat, kmem_protected_size,
and a series of kmemhighwater_* attributes should be set only by
or under the direction of kernel software developers or support
personnel.
See kmem_debug.
The maximum percentage of kernel virtual address space that is
reserved for kernel memory allocator (malloc) memory.
Default value: 25 (percent)
Minimum value: 0
Maximum value: 100
If this value is too low, no space in map errors may occur and
cause the system to hang. However, this situation indicates an
unexpected problem that should be reported to your support rep‐
resentative. Do not modify the default value for kmem_percent
unless directed to do so by your support representative.
See kmem_debug.
See kmem_debug.
See kmem_debug.
See kmem_debug.
See kmem_debug.
The percentage of memory that is reserved for the malloc compo‐
nent and used for kernel memory allocations that are less than
or equal to the page size (8 KB).
Default value: 0, which actually specifies 0.4 percent of avail‐
able memory or 256 KB, whichever is smaller.
Minimum value: 1
Maximum value: 75
A heavy network load that results in dropped packets is the pri‐
mary reason why you might want to increase the kmemreserve_per‐
cent value. The netstat -d -i or the vmstat -M command shows you
whether there are dropped packets. Increase the kmemreserve_per‐
cent value in small increments until there are no entries in the
fail_nowait column of the vmstat -M command's output display.
This value can be modified at run time. However, once the value
has been increased, you cannot decrease it below 1 percent at
run time. To restore the default setting (which is calculated by
the kernel), you must directly edit the /etc/sysconfigtab file
to remove or comment out the attribute and then reboot the sys‐
tem. This restriction will be removed in a future release.
(Under ordinary conditions, direct editing of the /etc/syscon‐
figtab file is a practice that is strongly discouraged.)
A setting that enables (1) or disables (0) a set of attribute
values that improve performance for 32-MB systems. This
attribute will soon be obsolete. Most kernel components now
determine memory size directly and apply more sophisticated
algorithms to determine defaults best suited for the system.
Default value: 1 if memory size is less than or equal to 32 MB
and 0 if memory size is greater than 32 MB
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
The full path to the directory where live dumps are written.
This attribute allows you to write dumps generated while the
system is running to a directory other than the one where crash
dumps are written.
Default value: /var/adm/crash (same directory as used for crash
dumps)
Minimum string length: 1 (byte)
Maximum string length: 65 (bytes)
This value can be modified at run time.
See the dump_savecnt entry for overview information about dump-
related attributes.
A value that enables (1) or disables (0) zero compression of
dumps that are generated while the system is still running. The
zero compression process produces files that take longer to cre‐
ate but occupy less space on disk.
Default value: 1 (enabled)
This value can be modified at run time. You might want to set
this value to 1 if you consistently use tools, such as those
from third parties, that do not support compressed dump files or
you want dumps created in a minimum amount of time. Because the
system is still running while a live dump is being created, some
of the data in a live dump may be inconsistent. The shorter cre‐
ation time may result in less data inconsistency.
See the dump_savecnt entry for overview information about dump-
related attributes. See the Kernel Debugging and the System
Administration manuals for more information about creating and
working with dump files.
A value that enables (1) or disables (0) lock debugging. If
enabled (1), the default value for the lockmode attribute is set
to 4.
If you explicitly specify a value for the lockmode attribute,
that value also causes lockdebug to be set or reset (overrid‐
den). For example, if you set lockmodeto be 0, 1, 2, or 3, the
value of lockdebug will be 0. If you set lockmode to be 4, the
value of lockdebug will be 1.
Default value: 0 (disabled)
Used for internal debugging.
Default value: 0
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
The mode in which the lock package, simple (spin) or queued, is
used within the kernel. (See the discussion of the locktype
attribute for information about the type of lock package used
for different systems.) Mode settings, which apply to both lock
packages, support different combinations of real-time (RT) ker‐
nel preemption, symmetric multiprocessing (SMP), and lock debug‐
ging with lock statistics: Neither RT nor SMP is required. In
this case, the calls to the lock primitives are patched out com‐
pletely. Only RT is required. In this case, the lock operations
maintain a “preemption blocking” count. Only SMP is required.
In this case, the lock operations provide synchronization among
multiple CPUs. Both RT and SMP are required. In this case,
lock operations are performed as described for both mode 1 and
2. Both RT and SMP are required. In this case, lock operations
are as described for mode 3, with the addition of kernel lock
debugging and statistics.
Default value: Assigned at boot time, depending on the values
for the rt_preempt_opt, cpu_enable_mask, and lockdebug
attributes, and on whether multiple CPUs are able to be booted.
If you specify a value for lockmode (0-4), your setting both
overrides the default setting and disables any of the three
related capabilities that cannot be supported by the chosen set
of lock primitives. For example, setting lockmode to 0, 1, 2, or
3 disables kernel lock debugging and statistics, whether or not
the lockdebug attribute was originally set to 1 (to enable ker‐
nel lock debugging and statistics).
The number of seconds that a CPU will wait on a lock. If a CPU
cannot acquire a lock in the specified amount of time, a fatal
error occurs, and the system panics and issues a message that
the lock time limit was exceeded.
Default value: 15 (seconds)
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit
instructions. A 0 value might cause boot failures on a multipro‐
cessor system. Values larger than the default might hang the
system instead of causing it to panic and reboot.
A value that specifies the type of locking package that the ker‐
nel software uses on the system. When the value is 0, simple
(spin) locking is used. When the value is 1, queued locking is
used. (A value of 1 is ignored on non-NUMA platforms.)
Default value: 1 on NUMA platforms; otherwise, 0.
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit
instructions.
The maximum number of bytes supported for a user's login name
string. One byte may be equivalent to one character in many lan‐
guages; however, byte-to-character equivalence varies, depending
on the user's default locale.
Default value: 12 (bytes)
Minimum value: 9
Maximum value: 64
In a TruCluster environment, the value of this attribute must be
the same on all member systems.
The default 12-byte limit exists only for backward compatibility
reasons. Increase login_name_max to 64 to enable longer login
names.
The depth to which complex locks can be nested for a thread at
one time. The value of max_lock_per_thread is used for debug‐
ging (when lockmode=4).
Default value: 32
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Cluster member ID that is set only by TruCluster software.
Default value: 0 (for standalone systems)
Minimum value: 0
Maximum value: 255
Reserved for future use by TruCluster software.
A kernel debugging value intended for internal use only.
A kernel debugging value intended for internal use only.
A kernel debugging value intended for internal use only.
A kernel debugging value intended for internal use only.
The size (in bytes) of the message buffer that is used to store
boot log messages.
Default value: Set automatically, based on memory size.
Minimum value: 0 (bytes)
Maximum value: 1,048,576
You may need to increase this value on a large system that
issues too many boot messages for the buffer to hold until the
syslog daemon starts. If the buffer is too small, initial boot
messages are not written to the system log and the beginning of
the first message that is written is likely to be truncated.
A value that is set and used internally to support a rolling
upgrade operation in a TruCluster environment. Do not modify
manually.
A value that is set and used internally to support a rolling
upgrade operation in a TruCluster environment. Do not modify
manually.
A value that changes the behavior of an internal system call.
Default value: 1 (on)
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation. Setting this value to off (0) has a negative impact
on system performance.
A value that is set and used internally to support a rolling
upgrade operation in a TruCluster environment. Do not modify
manually.
A value that is set and used internally to support a rolling
upgrade operation in a TruCluster environment. Do not modify
manually.
A value that determines whether the system generates partial (1)
or full (0) crash dumps when crash dumps are enabled.
Default value: 1 (partial dumps are created)
You can modify this attribute at run time. See the entry for the
dump_savecnt attribute for general information about dump-
related attributes.
A value that controls how I/O requests are gathered when readv
or writev operations are directed at a character device (not a
block device). For example, when set to 65536, up to 8 8-KB
buffers coalesce into one 64-KB buffer. This improves the effi‐
ciency of database operations.
Default value: 65,536
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
Reserved for future use; this value is not currently used by the
kernel.
A value that is set and used internally to support a rolling
upgrade operation in a TruCluster environment. Do not modify
manually.
A value that enables (1) or disables (0) real-time kernel pre‐
emption.
Default value: 0 (disabled)
If real-time kernel preemption is enabled (1): The default value
for lockmode will be 1 on a single-CPU system or 3 on a multi-
CPU system. However, if you explicitly specify certain lockmode
values, you implicitly override the rt_preempt_opt value. For
example, if you set lockmode to 0 or 2, you implicitly set
rt_preempt_opt to 0, disabling real-time kernel preemption. The
default value for ufs_lockholdmax will be 50. See
sys_attrs_ufs(5) for information on ufs_lockholdmax.
On a NUMA system (such as an ES80 or GS1280), the amount of time
(in seconds) that must elapse for a PAG to switch from a sender
to a receiver of threads (or the reverse).
Default value: 20 (seconds)
Minimum value: 0
Maximum value: 2,147,483,647
The maximum distance away from a home RAD that a remote RAD is
allowed to be for a thread to be scheduled on one of its proces‐
sors. This value applies to NUMA systems like the ES80 or
GS1280, on which RAD connections form a mesh, such that differ‐
ent RADs are different distances apart. The distance from one
RAD to other RADs is measured in terms of RAD-to-RAD hops in
different directions along the mesh. For example, setting
sched_distance to 3 means that a thread should be sheduled on a
RAD that is no more than two hops away in any direction from the
assigned home RAD.
Default value: 0 (hops)
Minimum value: 0
Maximum value: 63
An interval (in seconds) that sets the rate at which NUMA load-
balancing operations are done by the kernel.
Default value: 20 (seconds)
Minimum value: 1
Maximum value: 2,147,483,647
The minimum interval (in seconds) that must elapse before the
kernel can migrate a process or task to another RAD.
Default value: 30 (seconds)
Minimum value: 0
Maximum value: 2,147,483,647 A value used to enforce compliance
by system calls to the revised UNIX98 standards. If this
attribute is set to 0, then the system calls would exhibit their
default behavior. If it is set to 1, then all system calls will
behave in compliance with the revised UNIX98 standards. Stan‐
dard-compliant behavior for specific system calls can be turned
on using the following values. The strings in upper case are
macros defined in /usr/include/sys/param.h.
Default value: 0 - Existing behavior
STD_UNIX98_ALL: 1 - Standard compliance by all
UNIX98_POLLWRBAND: 2 - Switches poll() to standard
UNIX98_WAITPID: 4 - Switches waitpid() to standard
Note that all other system calls are compliant with the revised
UNIX98 standards by default and do not require setting this
attribute.
A value that restores (1) or corrects (0) backward-compatible
but incorrect behavior by the kernel's IEEE handler with respect
to fast mode floating-point exception traps.
When use_faulty_fpe_traps is set to 0, the IEEE handler cor‐
rectly returns a *_TRAP code for imprecise or invalid traps that
cannot be resolved and re-executed.
When use_faulty_fpe_traps is set to 1, the IEEE handler reverts
to returning a *_FAULT code for faulty traps that cannot be
fixed and re-executed. However, *_FAULT codes should be returned
only when the trap's trigger PC and instructions can still be
determined; otherwise, re-execution cannot be completed.
See ieee(3) and the Alpha Architecture Reference Manual for more
information on this topic.
Default value: 0 (correct return of *_TRAP code for faulty traps
that cannot be resolved and re-executed)
It is strongly recommended that you do not modify the default
value.
The CFG run-time points that are configurable from user mode.
Setting this attribute to a particular value triggers kernel
callbacks that have been registered for the point corresponding
to that value.
Default value: 45,000
Do not modify the default setting for this attribute unless
instructed to do so by support personnel or by patch kit docu‐
mentation.
The version string value (maximum length of 256 bytes) that is
set at kernel configuration time by the operating system soft‐
ware and returned by the sizer -v command. For example:
HP Tru64 UNIX V5.1B (Rev. 2650); Mon Oct 10 15:32:40 EST 2006
The string value (maximum length of 32 bytes) that is set at
kernel configuration time by the operating system software to
specify the abbreviated vendor name. For example:
HP
The string value (maximum length of 64 bytes) that is set at
kernel configuration time by the operating system software to
specify the product banner. For example:
HP Tru64 UNIX
The string value (maximum length of 32 bytes) that is set at
kernel configuration time by the operating system software to
specify the product name. For example:
Tru64 UNIX
The string value (maximum length of 64 bytes) that is set at
kernel configuration time by the operating system software to
specify the product release. For example:
5.1B (Rev. 2650)
The string value (maximum length of 128 bytes) that is set at
kernel configuration time by the operating system software to
specify the full vendor name. For example:
Hewlett-Packard Company
SEE ALSO
Commands: dxkerneltuner(8), sysconfig(8), sysconfigdb(8)
Others: sys_attrs(5)
System Configuration and Tuning
System Administration
Kernel Debugging
sys_attrs_generic(5)
