Manpages - lsfd.1

lsfd - list file descriptors

lsfd [option]

lsfd is intended to be a modern replacement for lsof*(8) on Linux systems. Unlike *lsof, lsfd is specialized to Linux kernel; it supports Linux specific features like namespaces with simpler code. lsfd is not a drop-in replacement for lsof; they are different in the command line interface and output formats.

The default output is subject to change. So whenever possible, you should avoid using default outputs in your scripts. Always explicitly define expected columns by using –output columns-list in environments where a stable output is required.

lsfd uses Libsmartcols for output formatting and filtering. See the description of –output option for customizing the output format, and –filter option for filtering. Use lsfd –help to get a list of all available columns.

-l, –threads

List in threads level.

-J, –json

Use JSON output format.

-n, –noheadings

Don't print headings.

-o, –output list

Specify which output columns to print. See the OUTPUT COLUMNS section for details of available columns.

The default list of columns may be extended if list is specified in the format +/list/ (e.g., lsfd -o +DELETED).

-r, –raw

Use raw output format.

–notruncate

Don't truncate text in columns.

-p, –pid pids

Collect information only for specified processes. pids is a list of pids. A comma or whitespaces can be used as separators. You can use this option with pidof*(1). See *FILTER EXAMPLES.

Both -Q option with an expression including PID, e.g. -Q (PID == 1), and -p option, e.g. -p 1, may print the same output but using -p option is much more efficient because -p option works at a much earlier stage of processing than the -Q option.

-i[4|6], –inet[=4|=6]

List only IPv4 sockets and/or IPv6 sockets.

-Q, –filter expr

Print only the files matching the condition represented by the expr. See also FILTER EXAMPLES.

-C, –counter label:/filter_expr/

Define a custom counter used in –summary output. lsfd makes a counter named label. During collect information, lsfd counts files matching filter_expr, and stores the counted number to the counter named label. lsfd applies filters defined with –filter options before counting; files excluded by the filters are not counted.

See FILTER EXPRESSION about filter_expr. label should not include { nor :. You can define multiple counters by specifying this option multiple times.

See also COUNTER EXAMPLES.

–summary[=/when/]

This option controls summary lines output. The optional argument when can be only, append or never. If the when argument is omitted, it defaults to only.

The summary reports counters. A counter consists of a label and an integer value. –counter is the option for defining a counter. If a user defines no counter, lsfd uses the definitions of pre-defined built-in counters (default counters) to make the summary output.

CAUTION: Using –summary and –json may make the output broken. Only combining –summary*=*only and –json is valid.

–debug-filter

Dump the internal data structure for the filter and exit. This is useful only for lsfd developers.

–dump-counters

Dump the definition of counters used in –summary output.

-h, –help

Display help text and exit.

-V, –version

Print version and exit.

Each column has a type. Types are surround by < and >.

CAUTION: The names and types of columns are not stable yet. They may be changed in the future releases.

AINODECLASS <=string=>

Class of anonymous inode.

ASSOC <=string=>

Association between file and process.

BLKDRV <=string=>

Block device driver name resolved by /proc/devices.

CHRDRV <=string=>

Character device driver name resolved by /proc/devices.

COMMAND <=string=>

Command of the process opening the file.

DELETED <=boolean=>

Reachability from the file system.

DEV <=string=>

ID of the device containing the file.

DEVTYPE <=string=>

Device type (blk, char, or nodev).

ENDPOINT <=string=>

IPC endpoints information communicated with the fd. The format of the column depends on the object associated with the fd:

FIFO type

PID,/COMMAND/,/ASSOC/[-r][-w]

The last characters ([-r][-w]) represents the read and/or write mode of the endpoint.

lsfd collects endpoints within the processes that lsfd scans; lsfd may miss some endpoints if you limits the processes with -p option.

FD <=number=>

File descriptor for the file.

FLAGS <=string=>

Flags specified when opening the file.

FUID <=number=>

User ID number of the file's owner.

INET.LADDR <=string=>

Local IP address.

INET.RADDR <=string=>

Remote IP address.

INET6.LADDR <=string=>

Local IP6 address.

INET6.RADDR <=string=>

Remote IP6 address.

INODE <=number=>

Inode number.

KNAME <=string=>

Raw file name extracted from from =/proc/=/pid/=/fd/=/fd/ or =/proc/=/pid/=/map_files/=/region/.

KTHREAD <=boolean=>

Whether the process is a kernel thread or not.

MAJ:MIN <=string=>

Device ID for special, or ID of device containing file.

MAPLEN <=number=>

Length of file mapping (in page).

MISCDEV <=string=>

Misc character device name resolved by /proc/misc.

MNTID <=number=>

Mount ID.

MODE <=string=>

Access mode (rwx).

NAME <=string=>

Cooked version of KNAME. It is mostly same as KNAME.

Some files have special formats and information sources:

NETLINK

protocol=/NETLINK.PROTOCOL/[ lport=/NETLINK.LPORT/[ group=/NETLINK.GROUPS/]]

PACKET

type=/SOCK.TYPE/[ protocol=/PACKET.PROTOCOL/][ iface=/PACKET.IFACE/]

pidfd

pid=/TARGET-PID/ comm=/TARGET-COMMAND/ nspid=/TARGET-NSPIDS/

lsfd extracts TARGET-PID and TARGET-NSPIDS from =/proc/=/pid/=/fdinfo/=/fd/.

PING

state=/SOCK.STATE/[ id=/PING.ID/][ laddr=/INET.LADDR/ [ raddr=/INET.RADDR/]]

PINGv6

state=/SOCK.STATE/[ id=/PING.ID/][ laddr=/INET6.LADDR/ [ raddr=/INET6.RADDR/]]

RAW

state=/SOCK.STATE/[ protocol=/RAW.PROTOCOL/ [ laddr=/INET.LADDR/ [ raddr=/INET.RADDR/]]]

RAWv6

state=/SOCK.STATE/[ protocol=/RAW.PROTOCOL/ [ laddr=/INET6.LADDR/ [ raddr=/INET6.RADDR/]]]

TCP, TCPv6

state=/SOCK.STATE/[ laddr=/TCP.LADDR/ [ raddr=/TCP.RADDR/]]

UDP, UDPv6

state=/SOCK.STATE/[ laddr=/UDP.LADDR/ [ raddr=/UDP.RADDR/]]

lsfd hides raddr= if UDP.RADDR is 0.0.0.0 and UDP.RPORT is 0.

UDP-LITE, UDPLITEv6

state=/SOCK.STATE/[ laddr=/UDPLITE.LADDR/ [ raddr=/UDPLITE.RADDR/]]

UNIX-STREAM

state=/SOCK.STATE/[ path=/UNIX.PATH/]

UNIX

state=/SOCK.STATE/[ path=/UNIX.PATH/] type=/SOCK.TYPE/

NETLINK.GROUPS <=number=>>

Netlink multicast groups.

NETLINK.LPORT <=number=>>

Netlink local port id.

NETLINK.PROTOCOL <=string=>>

Netlink protocol.

NLINK <=number=>

Link count.

NS.NAME <=string=>

Name (NS.TYPE:[/INODE/]) of the namespace specified with the file.

NS.TYPE <=string=>

Type of the namespace specified with the file. The type is mnt, cgroup, uts, ipc, user, pid, net, time, or unknown.

OWNER <=string=>

Owner of the file.

PACKET.IFACE <=string=>

Interface name associated with the packet socket.

PACKET.PROTOCOL <=string=>

L3 protocol associated with the packet socket.

PARTITION <=string=>

Block device name resolved by /proc/partition.

PID <=number=>

PID of the process opening the file.

PIDFD.COMM <=string=>

Command of the process targeted by the pidfd.

PIDFD.NSPID <=string=>

Value of NSpid field in =/proc/=/pid/=/fdinfo/=/fd/ of the pidfd.

Quoted from kernel/fork.c of Linux source tree:

If pid namespaces are supported then this function will also print the pid of a given pidfd refers to for all descendant pid namespaces starting from the current pid namespace of the instance, i.e. the Pid field and the first entry in the NSpid field will be identical.

Note that this differs from the Pid and NSpid fields in /proc/<pid>/status where Pid and NSpid are always shown relative to the pid namespace of the procfs instance.\\

PIDFD.PID <=number=>

PID of the process targeted by the pidfd.

PING.ID <`number`>

ICMP echo request id used on the PING socket.

POS <=number=>

File position.

RAW.PROTOCOL <=number=>

Protocol number of the raw socket.

RDEV <=string=>

Device ID (if special file).

SIZE <=number=>

File size.

SOCK.LISTENING <=boolean=>

Listening socket.

SOCK.NETS <=number=>

Inode identifying network namespace where the socket belongs to.

SOCK.PROTONAME <=string=>

Protocol name.

SOCK.STATE <=string=>

State of socket.

SOCK.TYPE <=string=>

Type of socket. Here type means the second parameter of socket system call:

·

stream

·

dgram

·

raw

·

rdm

·

seqpacket

·

dccp

·

packet

SOURCE <=string=>

File system, partition, or device containing the file.

STTYPE <=string=>

Raw file types returned from *stat*(2): BLK, CHR, DIR, FIFO, LINK, REG, SOCK, or UNKN.

TCP.LADDR <=string=>

Local L3 (INET.LADDR or INET6.LADDR) address and local TCP port.

TCP.LPORT <=integer=>

Local TCP port.

TCP.RADDR <=string=>

Remote L3 (INET.RADDR or INET6.RADDR) address and remote TCP port.

TCP.RPORT <=integer=>

Remote TCP port.

TID <=number=>

Thread ID of the process opening the file.

TYPE <=string=>

Cooked version of STTYPE. It is same as STTYPE with exceptions. For SOCK, print the value for SOCK.PROTONAME. For UNKN, print the value for AINODECLASS if SOURCE is anon_inodefs.

UDP.LADDR <=string=>

Local IP address and local UDP port.

UDP.LPORT <=integer=>

Local UDP port.

UDP.RADDR <=string=>

Remote IP address and remote UDP port.

UDP.RPORT <=integer=>

Remote UDP port.

UDPLITE.LADDR <=string=>

Local IP address and local UDPLite port.

UDPLITE.LPORT <=integer=>

Local UDP port.

UDPLITE.RADDR <=string=>

Remote IP address and remote UDPLite port.

UDPLITE.RPORT <=integer=>

Remote UDP port.

UID <=number=>

User ID number.

UNIX.PATH <=string=>

Filesystem pathname for UNIX domain socket.

USER <=string=>

User of the process.

lsfd evaluates the expression passed to –filter option every time before printing a file line. lsfd prints the line only if the result of evaluation is true.

An expression consists of column names, literals and, operators like: DELETED, (PID = 1)=, (NAME = "/etc/passwd")=, (PID = 1) && DELETED=. DELETED, PID, and NAME are column names in the example. 1 and "/etc/passwd" are literals. == and && are operators.

Before evaluation, lsfd substitutes column names in the given expression with actual column values in the line. There are three different data types: boolean, string, and number. For columns with a boolean type, the value can be stand-alone. For string and number values, the value must be an operand of an operator, for example, (PID = 1)=. See OUTPUT COLUMNS about the types of columns.

Literal is for representing a value directly. See BOOLLIT, STRLIT, and NUMLIT. Different data types have different literal syntax.

An operator works with one or two operand(s). An operator has an expectation about the data type(s) of its operands. Giving an unexpected data type to an operator causes a syntax error.

Operators taking two operands are and, or, eq, ne, le, lt, ge, gt, =~, !~. Alphabetically named operators have C-language flavored aliases: &&, ||, ==, !=, <, ⇐, >=, and >.

! is the only operator that takes one operand.

eq, ne, and their aliases expect operands have the same data type. Applying these operators return a boolean.

and, or, not and their aliases expect operands have boolean data type. Applying these operators return a boolean.

lt, le, gt, ge, and their aliases expect operands have number data types. Applying these operators return a boolean.

=~ is for regular expression matching; if a string at the right side matches a regular expression at the left side, the result is true. The right side operand must be a string literal. See STRLIT about the syntax.

!~ is a short-hand version of not (STR =~ PAT); it inverts the result of =~.

lsfd has few options for filtering. In most of cases, what you should know is -Q (or –filter) option. Combined with -o (or –output) option, you can customize the output as you want.

List files associated with PID 1 and PID 2 processes:

# lsfd -Q (PID == 1) or (PID == 2)

Do the same in an alternative way:

# lsfd -Q (PID == 1) || (PID == 2)

Do the same in a more efficient way:

# lsfd --pid 1,2

Whitescapes can be used instead of a comma:

# lsfd --pid 1 2

Utilize *pidof*(1) for list the files associated with "firefox":

# lsfd --pid "$(pidof firefox)"

List the 1st file descriptor opened by PID 1 process:

# lsfd -Q (PID == 1) and (FD == 1)

Do the same in an alternative way:

# lsfd -Q (PID == 1) && (FD == 1)

List all running executables:

# lsfd -Q ASSOC == "exe"

Do the same in an alternative way:

# lsfd -Q ASSOC eq "exe"

Do the same but print only file names:

# lsfd -o NAME -Q ASSOC eq "exe" | sort -u

List deleted files associated to processes:

# lsfd -Q DELETED

List non-regular files:

# lsfd -Q TYPE != "REG"

List block devices:

# lsfd -Q DEVTYPE == "blk"

Do the same with TYPE column:

# lsfd -Q TYPE == "BLK"

List files including "dconf" directory in their names:

# lsfd -Q NAME =~ ".\*/dconf/.*"

List files opened in a QEMU virtual machine:

# lsfd -Q (COMMAND =~ ".\*qemu.*") and (FD >= 0)

Hide files associated to kernel threads:

# lsfd -Q !KTHREAD

Report the numbers of netlink socket descriptors and unix socket descriptors:

# lsfd --summary=only \
        -C netlink sockets:(NAME =~ "NETLINK:.*") \
        -C unix sockets:(NAME =~ "UNIX:.*")
VALUE COUNTER
   57 netlink sockets
 1552 unix sockets

Do the same but print in JSON format:

# lsfd --summary=only --json \
        -C netlink sockets:(NAME =~ "NETLINK:.*") \
        -C unix sockets:(NAME =~ "UNIX:.*")
{
   "lsfd-summary": [
      {
         "value": 15,
         "counter": "netlink sockets"
      },{
         "value": 798,
         "counter": "unix sockets"
      }
   ]
}

The lsfd command is part of the util-linux package since v2.38.

*lsof*(8) *pidof*(1) *proc*(5) *socket*(2) *stat*(2)

For bug reports, use the issue tracker at <https://github.com/util-linux/util-linux/issues>.

The lsfd command is part of the util-linux package which can be downloaded from Linux Kernel Archive <https://www.kernel.org/pub/linux/utils/util-linux/>.