1.\" Copyright (c) 1991 The Regents of the University of California. 2.\" All rights reserved. 3.\" 4.\" Redistribution and use in source and binary forms, with or without 5.\" modification, are permitted provided that the following conditions 6.\" are met: 7.\" 1. Redistributions of source code must retain the above copyright 8.\" notice, this list of conditions and the following disclaimer. 9.\" 2. Redistributions in binary form must reproduce the above copyright 10.\" notice, this list of conditions and the following disclaimer in the 11.\" documentation and/or other materials provided with the distribution. 12.\" 3. Neither the name of the University nor the names of its contributors 13.\" may be used to endorse or promote products derived from this software 14.\" without specific prior written permission. 15.\" 16.\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 17.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 18.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 19.\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 20.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 21.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 22.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 23.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 24.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 25.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 26.\" SUCH DAMAGE. 27.\" 28.Dd March 24, 2024 29.Dt MEM 4 30.Os 31.Sh NAME 32.Nm mem , 33.Nm kmem 34.Nd memory files 35.Sh SYNOPSIS 36.Cd "device mem" 37.Sh DESCRIPTION 38The special file 39.Pa /dev/mem 40is an interface to the physical memory of the computer. 41Byte offsets in this file are interpreted as physical memory addresses. 42Reading and writing this file is equivalent to reading and writing 43memory itself. 44Only offsets within the bounds of 45.Pa /dev/mem 46are allowed. 47.Pp 48Kernel virtual memory is accessed through the interface 49.Pa /dev/kmem 50in the same manner as 51.Pa /dev/mem . 52Only kernel virtual addresses that are currently mapped to memory are allowed. 53.Pp 54On ISA the I/O memory space begins at physical address 0x000a0000 55and runs to 0x00100000. 56The 57per-process data 58size 59for the current process 60is 61.Dv UPAGES 62long, and ends at virtual 63address 0xf0000000. 64.Sh IOCTL INTERFACE 65.Ss Address Properties 66The 67.Dv MEM_EXTRACT_PADDR 68ioctl can be used to look up the physical address and NUMA domain of a given 69virtual address in the calling process' address space. 70The request is described by 71.Bd -literal 72struct mem_extract { 73 uint64_t me_vaddr; /* input */ 74 uint64_t me_paddr; /* output */ 75 int me_domain; /* output */ 76 int me_state; /* output */ 77}; 78.Ed 79.Pp 80The ioctl returns an error if the address is not valid. 81The information returned by 82.Dv MEM_EXTRACT_PADDR 83may be out of date by the time that the ioctl call returns. 84Specifically, concurrent system calls, page faults, or system page reclamation 85activity may have unmapped the virtual page or replaced the backing physical 86page before the ioctl call returns. 87Wired pages, e.g., those locked by 88.Xr mlock 2 , 89will not be reclaimed by the system. 90.Pp 91The 92.Fa me_state 93field provides information about the state of the virtual page: 94.Bl -tag -width indent 95.It Dv ME_STATE_INVALID 96The virtual address is invalid. 97.It Dv ME_STATE_VALID 98The virtual address is valid but is not mapped at the time of the ioctl call. 99.It Dv ME_STATE_MAPPED 100The virtual address corresponds to a physical page mapping, and the 101.Fa me_paddr 102and 103.Fa me_domain 104fields are valid. 105.El 106.Ss Memory Ranges 107.Pp 108Several architectures allow attributes to be associated with ranges of physical 109memory. 110These attributes can be manipulated via 111.Fn ioctl 112calls performed on 113.Pa /dev/mem . 114Declarations and data types are to be found in 115.In sys/memrange.h . 116.Pp 117The specific attributes, and number of programmable ranges may vary between 118architectures. 119The full set of supported attributes is: 120.Bl -tag -width indent 121.It Dv MDF_UNCACHEABLE 122The region is not cached. 123.It Dv MDF_WRITECOMBINE 124Writes to the region may be combined or performed out of order. 125.It Dv MDF_WRITETHROUGH 126Writes to the region are committed synchronously. 127.It Dv MDF_WRITEBACK 128Writes to the region are committed asynchronously. 129.It Dv MDF_WRITEPROTECT 130The region cannot be written to. 131.El 132.Pp 133Memory ranges are described by 134.Bd -literal 135struct mem_range_desc { 136 uint64_t mr_base; /* physical base address */ 137 uint64_t mr_len; /* physical length of region */ 138 int mr_flags; /* attributes of region */ 139 char mr_owner[8]; 140}; 141.Ed 142.Pp 143In addition to the region attributes listed above, the following flags 144may also be set in the 145.Fa mr_flags 146field: 147.Bl -tag -width indent 148.It MDF_FIXBASE 149The region's base address cannot be changed. 150.It MDF_FIXLEN 151The region's length cannot be changed. 152.It MDF_FIRMWARE 153The region is believed to have been established by the system firmware. 154.It MDF_ACTIVE 155The region is currently active. 156.It MDF_BOGUS 157We believe the region to be invalid or otherwise erroneous. 158.It MDF_FIXACTIVE 159The region cannot be disabled. 160.It MDF_BUSY 161The region is currently owned by another process and may not be 162altered. 163.El 164.Pp 165Operations are performed using 166.Bd -literal 167struct mem_range_op { 168 struct mem_range_desc *mo_desc; 169 int mo_arg[2]; 170}; 171.Ed 172.Pp 173The 174.Dv MEMRANGE_GET 175ioctl is used to retrieve current memory range attributes. 176If 177.Va mo_arg[0] 178is set to 0, it will be updated with the total number of memory range 179descriptors. 180If greater than 0, the array at 181.Va mo_desc 182will be filled with a corresponding number of descriptor structures, 183or the maximum, whichever is less. 184.Pp 185The 186.Dv MEMRANGE_SET 187ioctl is used to add, alter and remove memory range attributes. 188A range 189with the 190.Dv MDF_FIXACTIVE 191flag may not be removed; a range with the 192.Dv MDF_BUSY 193flag may not be removed or updated. 194.Pp 195.Va mo_arg[0] 196should be set to 197.Dv MEMRANGE_SET_UPDATE 198to update an existing or establish a new range, or to 199.Dv MEMRANGE_SET_REMOVE 200to remove a range. 201.El 202.Ss Live Kernel Dumps 203.Pp 204The 205.Dv MEM_KERNELDUMP 206ioctl will initiate a kernel dump against the running system, the contents of 207which will be written to a process-owned file descriptor. 208The resulting dump output will be in minidump format. 209The request is described by 210.Bd -literal 211struct mem_livedump_arg { 212 int fd; /* input */ 213 int flags /* input */ 214 uint8_t compression /* input */ 215}; 216.Ed 217.Pp 218The 219.Va fd 220field is used to pass the file descriptor. 221.Pp 222The 223.Va flags 224field is currently unused and must be set to zero. 225.Pp 226The 227.Va compression 228field can be used to specify the desired compression to 229be applied to the dump output. 230The supported values are defined in 231.In sys/kerneldump.h ; 232that is, 233.Dv KERNELDUMP_COMP_NONE , 234.Dv KERNELDUMP_COMP_GZIP , 235or 236.Dv KERNELDUMP_COMP_ZSTD . 237.Pp 238Kernel dumps taken against the running system may have inconsistent kernel data 239structures due to allocation, deallocation, or modification of memory 240concurrent to the dump procedure. 241Thus, the resulting core dump is not guaranteed to be usable. 242A system under load is more likely to produce an inconsistent result. 243Despite this, live kernel dumps can be useful for offline debugging of certain 244types of kernel bugs, such as deadlocks, or in inspecting a particular part of 245the system's state. 246.Sh RETURN VALUES 247.Ss MEM_EXTRACT_PADDR 248The 249.Dv MEM_EXTRACT_PADDR 250ioctl always returns a value of zero. 251.Ss MEMRANGE_GET/MEMRANGE_SET 252.Bl -tag -width Er 253.It Bq Er EOPNOTSUPP 254Memory range operations are not supported on this architecture. 255.It Bq Er ENXIO 256No memory range descriptors are available (e.g., firmware has not enabled 257any). 258.It Bq Er EINVAL 259The memory range supplied as an argument is invalid or overlaps another 260range in a fashion not supported by this architecture. 261.It Bq Er EBUSY 262An attempt to remove or update a range failed because the range is busy. 263.It Bq Er ENOSPC 264An attempt to create a new range failed due to a shortage of hardware 265resources (e.g., descriptor slots). 266.It Bq Er ENOENT 267An attempt to remove a range failed because no range matches the descriptor 268base/length supplied. 269.It Bq Er EPERM 270An attempt to remove a range failed because the range is permanently 271enabled. 272.El 273.Ss MEM_KERNELDUMP 274.Bl -tag -width Er 275.It Bq Er EOPNOTSUPP 276Kernel minidumps are not supported on this architecture. 277.It Bq Er EPERM 278An attempt to begin the kernel dump failed because the calling thread lacks the 279.It Bq Er EBADF 280The supplied file descriptor was invalid, or does not have write permission. 281.It Bq Er EBUSY 282An attempt to begin the kernel dump failed because one is already in progress. 283.It Bq Er EINVAL 284An invalid or unsupported value was specified in 285.Va flags . 286.It Bq Er EINVAL 287An invalid or unsupported compression type was specified. 288.Dv PRIV_KMEM_READ 289privilege. 290.El 291.Sh FILES 292.Bl -tag -width /dev/kmem -compact 293.It Pa /dev/mem 294.It Pa /dev/kmem 295.El 296.Sh SEE ALSO 297.Xr kvm 3 , 298.Xr memcontrol 8 299.Sh HISTORY 300The 301.Pa /dev/mem 302file appeared in 303.At v1 304and 305.Pa /dev/kmem 306in 307.At v5 . 308The ioctl interface for memory range attributes was added in 309.Fx 3.2 . 310.Sh BUGS 311Busy range attributes are not yet managed correctly. 312.Pp 313This device is required for all users of 314.Xr kvm 3 315to operate. 316