diff --git a/share/man/man4/ddb.4 b/share/man/man4/ddb.4 index 08e804c64ff8..031df9c27bf2 100644 --- a/share/man/man4/ddb.4 +++ b/share/man/man4/ddb.4 @@ -1,1683 +1,1689 @@ .\" .\" Mach Operating System .\" Copyright (c) 1991,1990 Carnegie Mellon University .\" Copyright (c) 2007 Robert N. M. Watson .\" All Rights Reserved. .\" .\" Permission to use, copy, modify and distribute this software and its .\" documentation is hereby granted, provided that both the copyright .\" notice and this permission notice appear in all copies of the .\" software, derivative works or modified versions, and any portions .\" thereof, and that both notices appear in supporting documentation. .\" .\" CARNEGIE MELLON ALLOWS FREE USE OF THIS SOFTWARE IN ITS "AS IS" .\" CONDITION. CARNEGIE MELLON DISCLAIMS ANY LIABILITY OF ANY KIND FOR .\" ANY DAMAGES WHATSOEVER RESULTING FROM THE USE OF THIS SOFTWARE. .\" .\" Carnegie Mellon requests users of this software to return to .\" .\" Software Distribution Coordinator or Software.Distribution@CS.CMU.EDU .\" School of Computer Science .\" Carnegie Mellon University .\" Pittsburgh PA 15213-3890 .\" .\" any improvements or extensions that they make and grant Carnegie Mellon .\" the rights to redistribute these changes. .\" -.Dd June 10, 2025 +.Dd October 31, 2025 .Dt DDB 4 .Os .Sh NAME .Nm ddb .Nd interactive kernel debugger .Sh SYNOPSIS In order to enable kernel debugging facilities include: .Bd -ragged -offset indent .Cd options KDB .Cd options DDB .Ed .Pp To prevent activation of the debugger on kernel .Xr panic 9 : .Bd -ragged -offset indent .Cd options KDB_UNATTENDED .Ed .Pp In order to print a stack trace of the current thread on the console for a panic: .Bd -ragged -offset indent .Cd options KDB_TRACE .Ed .Pp To print the numerical value of symbols in addition to the symbolic representation, define: .Bd -ragged -offset indent .Cd options DDB_NUMSYM .Ed .Pp To enable the .Xr gdb 4 backend, so that remote debugging with .Xr kgdb 1 Pq Pa ports/devel/gdb is possible, include: .Bd -ragged -offset indent .Cd options GDB .Ed .Sh DESCRIPTION The .Nm kernel debugger is an interactive debugger with a syntax inspired by .Xr gdb 1 Pq Pa ports/devel/gdb . If linked into the running kernel, it can be invoked locally with the .Ql debug .Xr keymap 5 action, usually mapped to Ctrl+Alt+Esc, or by setting the .Va debug.kdb.enter sysctl to 1. The debugger is also invoked on kernel .Xr panic 9 if the .Va debug.debugger_on_panic .Xr sysctl 8 MIB variable is set non-zero, which is the default unless the .Dv KDB_UNATTENDED option is specified. Similarly, if the .Va debug.debugger_on_recursive_panic variable is set to .Dv 1 , then the debugger will be invoked on a recursive kernel panic. This variable has a default value of .Dv 0 , and has no effect if .Va debug.debugger_on_panic is already set non-zero. .Pp The current location is called .Va dot . The .Va dot is displayed with a hexadecimal format at a prompt. The commands .Ic examine and .Ic write update .Va dot to the address of the last line examined or the last location modified, and set .Va next to the address of the next location to be examined or changed. Other commands do not change .Va dot , and set .Va next to be the same as .Va dot . .Pp The general command syntax is: .Ar command Ns Op Li / Ns Ar modifier .Oo Ar addr Oc Ns Op , Ns Ar count .Pp A blank line repeats the previous command from the address .Va next with count 1 and no modifiers. Specifying .Ar addr sets .Va dot to the address. Omitting .Ar addr uses .Va dot . A missing .Ar count is taken to be 1 for printing commands or infinity for stack traces. A .Ar count of -1 is equivalent to a missing .Ar count . Options that are supplied but not supported by the given .Ar command are usually ignored. .Pp The .Nm debugger has a pager feature (like the .Xr more 1 command) for the output. If an output line exceeds the number set in the .Va lines variable, it displays .Dq Li --More-- and waits for a response. The valid responses for it are: .Pp .Bl -tag -compact -width ".Li SPC" .It Li SPC one more page .It Li RET one more line .It Li q abort the current command, and return to the command input mode .El .Pp Finally, .Nm provides a small (currently 10 items) command history, and offers simple .Nm emacs Ns -style command line editing capabilities. In addition to the .Nm emacs control keys, the usual ANSI arrow keys may be used to browse through the history buffer, and move the cursor within the current line. .Sh COMMANDS .Ss COMMON DEBUGGER COMMANDS .Bl -tag -width indent -compact .It Ic help Print a short summary of the available commands and command abbreviations. .Pp .It Xo .Ic examine Ns Op Li / Ns Cm AISabcdghilmorsuxz ... .Oo Ar addr Oc Ns Op , Ns Ar count .Xc .It Xo .Ic x Ns Op Li / Ns Cm AISabcdghilmorsuxz ... .Oo Ar addr Oc Ns Op , Ns Ar count .Xc Display the addressed locations according to the formats in the modifier. Multiple modifier formats display multiple locations. If no format is specified, the last format specified for this command is used. .Pp The format characters are: .Bl -tag -compact -width indent .It Cm b look at by bytes (8 bits) .It Cm h look at by half words (16 bits) .It Cm l look at by long words (32 bits) .It Cm g look at by quad words (64 bits) .It Cm a print the location being displayed .It Cm A print the location with a line number if possible .It Cm x display in unsigned hex .It Cm z display in signed hex .It Cm o display in unsigned octal .It Cm d display in signed decimal .It Cm u display in unsigned decimal .It Cm r display in current radix, signed .It Cm c display low 8 bits as a character. Non-printing characters are displayed as an octal escape code (e.g., .Ql \e000 ) . .It Cm s display the null-terminated string at the location. Non-printing characters are displayed as octal escapes. .It Cm m display in unsigned hex with character dump at the end of each line. The location is also displayed in hex at the beginning of each line. .It Cm i display as a disassembled instruction .It Cm I display as a disassembled instruction with possible alternate formats depending on the machine. On i386, this selects the alternate format for the instruction decoding (16 bits in a 32-bit code segment and vice versa). .It Cm S display a symbol name for the pointer stored at the address .El .Pp .It Ic xf Examine forward: execute an .Ic examine command with the last specified parameters to it except that the next address displayed by it is used as the start address. .Pp .It Ic xb Examine backward: execute an .Ic examine command with the last specified parameters to it except that the last start address subtracted by the size displayed by it is used as the start address. .Pp .It Ic print Ns Op Li / Ns Cm acdoruxz .It Ic p Ns Op Li / Ns Cm acdoruxz Print .Ar addr Ns s according to the modifier character (as described above for .Cm examine ) . Valid formats are: .Cm a , x , z , o , d , u , r , and .Cm c . If no modifier is specified, the last one specified to it is used. The argument .Ar addr can be a string, in which case it is printed as it is. For example: .Bd -literal -offset indent print/x "eax = " $eax "\enecx = " $ecx "\en" .Ed .Pp will print like: .Bd -literal -offset indent eax = xxxxxx ecx = yyyyyy .Ed .Pp .It Xo .Ic write Ns Op Li / Ns Cm bhl .Ar addr expr1 Op Ar expr2 ... .Xc .It Xo .Ic w Ns Op Li / Ns Cm bhl .Ar addr expr1 Op Ar expr2 ... .Xc Write the expressions specified after .Ar addr on the command line at succeeding locations starting with .Ar addr . The write unit size can be specified in the modifier with a letter .Cm b (byte), .Cm h (half word) or .Cm l (long word) respectively. If omitted, long word is assumed. .Pp .Sy Warning : since there is no delimiter between expressions, strange things may happen. It is best to enclose each expression in parentheses. .Pp .It Ic set Li $ Ns Ar variable Oo Li = Oc Ar expr Set the named variable or register with the value of .Ar expr . Valid variable names are described below. .Pp .It Ic break Ns Oo Li / Ns Cm u Oc Oo Ar addr Oc Ns Op , Ns Ar count .It Ic b Ns Oo Li / Ns Cm u Oc Oo Ar addr Oc Ns Op , Ns Ar count Set a break point at .Ar addr . If .Ar count is supplied, the .Ic continue command will not stop at this break point on the first .Ar count \- 1 times that it is hit. If the break point is set, a break point number is printed with .Ql # . This number can be used in deleting the break point or adding conditions to it. .Pp If the .Cm u modifier is specified, this command sets a break point in user address space. Without the .Cm u option, the address is considered to be in the kernel space, and a wrong space address is rejected with an error message. This modifier can be used only if it is supported by machine dependent routines. .Pp .Sy Warning : If a user text is shadowed by a normal user space debugger, user space break points may not work correctly. Setting a break point at the low-level code paths may also cause strange behavior. .Pp .It Ic delete Op Ar addr .It Ic d Op Ar addr .It Ic delete Li # Ns Ar number .It Ic d Li # Ns Ar number Delete the specified break point. The break point can be specified by a break point number with .Ql # , or by using the same .Ar addr specified in the original .Ic break command, or by omitting .Ar addr to get the default address of .Va dot . .Pp .It Ic halt Halt the system. .Pp .It Ic watch Oo Ar addr Oc Ns Op , Ns Ar size Set a watchpoint for a region. Execution stops when an attempt to modify the region occurs. The .Ar size argument defaults to 4. If you specify a wrong space address, the request is rejected with an error message. .Pp .Sy Warning : Attempts to watch wired kernel memory may cause unrecoverable error in some systems such as i386. Watchpoints on user addresses work best. .Pp .It Ic hwatch Oo Ar addr Oc Ns Op , Ns Ar size Set a hardware watchpoint for a region if supported by the architecture. Execution stops when an attempt to modify the region occurs. The .Ar size argument defaults to 4. .Pp .Sy Warning : The hardware debug facilities do not have a concept of separate address spaces like the watch command does. Use .Ic hwatch for setting watchpoints on kernel address locations only, and avoid its use on user mode address spaces. .Pp .It Ic dhwatch Oo Ar addr Oc Ns Op , Ns Ar size Delete specified hardware watchpoint. .Pp .It Ic kill Ar sig pid Send signal .Ar sig to process .Ar pid . The signal is acted on upon returning from the debugger. This command can be used to kill a process causing resource contention in the case of a hung system. See .Xr signal 3 for a list of signals. Note that the arguments are reversed relative to .Xr kill 2 . .Pp .It Ic step Ns Oo Li / Ns Cm p Oc Ns Op , Ns Ar count .It Ic s Ns Oo Li / Ns Cm p Oc Ns Op , Ns Ar count Single step .Ar count times. If the .Cm p modifier is specified, print each instruction at each step. Otherwise, only print the last instruction. .Pp .Sy Warning : depending on machine type, it may not be possible to single-step through some low-level code paths or user space code. On machines with software-emulated single-stepping (e.g., pmax), stepping through code executed by interrupt handlers will probably do the wrong thing. .Pp .It Ic continue Ns Op Li / Ns Cm c .It Ic c Ns Op Li / Ns Cm c Continue execution until a breakpoint or watchpoint. If the .Cm c modifier is specified, count instructions while executing. Some machines (e.g., pmax) also count loads and stores. .Pp .Sy Warning : when counting, the debugger is really silently single-stepping. This means that single-stepping on low-level code may cause strange behavior. .Pp .It Ic until Ns Op Li / Ns Cm p Stop at the next call or return instruction. If the .Cm p modifier is specified, print the call nesting depth and the cumulative instruction count at each call or return. Otherwise, only print when the matching return is hit. .Pp .It Ic next Ns Op Li / Ns Cm p .It Ic match Ns Op Li / Ns Cm p Stop at the matching return instruction. If the .Cm p modifier is specified, print the call nesting depth and the cumulative instruction count at each call or return. Otherwise, only print when the matching return is hit. .Pp .It Xo .Ic trace Ns Op Li / Ns Cm u .Op Ar pid | tid Ns .Op , Ns Ar count .Xc .It Xo .Ic t Ns Op Li / Ns Cm u .Op Ar pid | tid Ns .Op , Ns Ar count .Xc .It Xo .Ic where Ns Op Li / Ns Cm u .Op Ar pid | tid Ns .Op , Ns Ar count .Xc .It Xo .Ic bt Ns Op Li / Ns Cm u .Op Ar pid | tid Ns .Op , Ns Ar count .Xc Stack trace. The .Cm u option traces user space; if omitted, .Ic trace only traces kernel space. The optional argument .Ar count is the number of frames to be traced. If .Ar count is omitted, all frames are printed. .Pp .Sy Warning : User space stack trace is valid only if the machine dependent code supports it. .Pp .It Xo .Ic search Ns Op Li / Ns Cm bhl .Ar addr .Ar value .Op Ar mask Ns .Op , Ns Ar count .Xc Search memory for .Ar value . The optional .Ar count argument limits the search. .\" .Pp .It Xo .Ic Ic reboot Ns Op Li / Ns Cm s .Op Ar seconds .Xc .It Xo .Ic Ic reset Ns Op Li / Ns Cm s .Op Ar seconds .Xc Hard reset the system. If the optional argument .Ar seconds is given, the debugger will wait for this long, at most a week, before rebooting. When the .Cm s modifier is given, the command will skip running any registered shutdown handlers and attempt the most basic reset. .Pp .It Ic thread Ar addr | tid Switch the debugger to the thread with ID .Ar tid , if the argument is a decimal number, or address .Ar addr , otherwise. .Pp .It Ic watchdog Op Ar exp Program the .Xr watchdog 4 timer to fire in .Pf 2^ Ar exp seconds. If no argument is provided, the watchdog timer is disabled. .El .Ss SPECIALIZED HELPER COMMANDS .Bl -tag -width indent -compact .It Xo .Ic findstack .Ar addr .Xc Prints the address of the thread whose kernel-mode stack contains .Ar addr , if any. .Pp .It Ic show Cm active trace .It Ic acttrace Show a stack trace for every thread running on a CPU. .Pp .It Ic show Cm all procs Ns Op Li / Ns Cm a .It Ic ps Ns Op Li / Ns Cm a Display all process information. The process information may not be shown if it is not supported in the machine, or the bottom of the stack of the target process is not in the main memory at that time. The .Cm a modifier will print command line arguments for each process. .\" .Pp -.It Ic show Cm all tcpcbs Ns Op Li / Ns Cm b Ns Cm l +.It Ic show Cm all tcpcbs Ns Op Li / Ns Cm b Ns Cm i Ns Cm l Show the same output as "show tcpcb" does, but for all TCP control blocks within the system. The .Cm b modifier will request BBLog entries to be printed. +If the +.Cm i +modifier is provided, the corresponding IP control block is also shown. Using the .Cm l modifier will limit the output to TCP control blocks, which are locked. .\" .Pp .It Ic show Cm all trace .It Ic alltrace Show a stack trace for every thread in the system. .Pp .It Ic show Cm all ttys Show all TTY's within the system. Output is similar to .Xr pstat 8 , but also includes the address of the TTY structure. .\" .Pp .It Ic show Cm all vnets Show the same output as "show vnet" does, but lists all virtualized network stacks within the system. .\" .Pp .It Ic show Cm allchains Show the same information like "show lockchain" does, but for every thread in the system. .\" .Pp .It Ic show Cm alllocks Show all locks that are currently held. This command is only available if .Xr witness 4 is included in the kernel. .\" .Pp .It Ic show Cm allpcpu The same as "show pcpu", but for every CPU present in the system. .\" .Pp .It Ic show Cm allrman Show information related with resource management, including interrupt request lines, DMA request lines, I/O ports, I/O memory addresses, and Resource IDs. .\" .Pp .It Ic show Cm apic Dump data about APIC IDT vector mappings. .\" .Pp .It Ic show Cm badstacks Walk the .Xr witness 4 graph and print any lock-order violations. This command is only available if .Xr witness 4 is included in the kernel. .\" .Pp .It Ic show Cm breaks Show breakpoints set with the "break" command. .\" .Pp .It Ic show Cm bio Ar addr Show information about the bio structure .Vt struct bio present at .Ar addr . See the .Pa sys/bio.h header file and .Xr g_bio 9 for more details on the exact meaning of the structure fields. .\" .Pp .It Ic show Cm buffer Ar addr Show information about the buf structure .Vt struct buf present at .Ar addr . See the .Pa sys/buf.h header file for more details on the exact meaning of the structure fields. .\" .Pp .It Ic show Cm callout Ar addr Show information about the callout structure .Vt struct callout present at .Ar addr . .\" .Pp .It Ic show Cm cdev Op Ar addr Show the internal devfs state of the cdev structure located at .Ar addr . If no argument is provided, show the list of all created cdevs, consisting of the devfs node name and the .Vt struct cdev address. .\" .Pp .It Ic show Cm conifhk Lists hooks currently waiting for completion in .Fn run_interrupt_driven_config_hooks . .\" .Pp .It Ic show Cm cpusets Print numbered root and assigned CPU affinity sets. See .Xr cpuset 2 for more details. .\" .Pp .It Ic show Cm cyrixreg Show registers specific to the Cyrix processor. .\" .Pp .It Ic show Cm devmap Prints the contents of the static device mapping table. Currently only available on the ARM architecture. .\" .Pp .It Ic show Cm domain Ar addr Print protocol domain structure .Vt struct domain at address .Ar addr . See the .Pa sys/domain.h header file for more details on the exact meaning of the structure fields. .\" .Pp .It Ic show Cm ffs Op Ar addr Show brief information about ffs mount at the address .Ar addr , if argument is given. Otherwise, provides the summary about each ffs mount. .\" .Pp .It Ic show Cm file Ar addr Show information about the file structure .Vt struct file present at address .Ar addr . .\" .Pp .It Ic show Cm files Show information about every file structure in the system. .\" .Pp .It Ic show Cm freepages Show the number of physical pages in each of the free lists. .\" .Pp .It Ic show Cm geom Op Ar addr If the .Ar addr argument is not given, displays the entire GEOM topology. If .Ar addr is given, displays details about the given GEOM object (class, geom, provider or consumer). .\" .Pp .It Ic show Cm idt Show IDT layout. The first column specifies the IDT vector. The second one is the name of the interrupt/trap handler. Those functions are machine dependent. .\" .Pp .It Ic show Cm igi_list Ar addr Show information about the IGMP structure .Vt struct igmp_ifsoftc present at .Ar addr . .\" .Pp .It Ic show Cm iosched Ar addr Show information about the I/O scheduler .Vt struct cam_iosched_softc located at .Ar addr . .\" .Pp .It Ic show Cm inodedeps Op Ar addr Show brief information about each inodedep structure. If .Ar addr is given, only inodedeps belonging to the fs located at the supplied address are shown. .\" .Pp .It Ic show Cm inpcb Ar addr Show information on IP Control Block .Vt struct in_pcb present at .Ar addr . .\" .Pp .It Ic show Cm intr Dump information about interrupt handlers. .\" .Pp .It Ic show Cm intrcnt Dump the interrupt statistics. .\" .Pp .It Ic show Cm irqs Show interrupt lines and their respective kernel threads. .\" .Pp .It Ic show Cm ktr Ns Op Li / Ns Cm a Ns Cm v Ns Cm V Print the contents of the .Xr ktr 4 trace buffer. The .Cm v modifier will request fully verbose output, causing the file, line number, and timestamp to be printed for each trace entry. The .Cm V modifier will request only the timestamps to be printed. The .Cm a modifier will request that the output be unpaginated. .\" .Pp .It Ic show Cm lapic Show information from the local APIC registers for this CPU. .\" .Pp .It Ic show Cm lock Ar addr Show lock structure. The output format is as follows: .Bl -tag -width "flags" .It Ic class : Class of the lock. Possible types include .Xr mutex 9 , .Xr rmlock 9 , .Xr rwlock 9 , .Xr sx 9 . .It Ic name : Name of the lock. .It Ic flags : Flags passed to the lock initialization function. .Em flags values are lock class specific. .It Ic state : Current state of a lock. .Em state values are lock class specific. .It Ic owner : Lock owner. .El .\" .Pp .It Ic show Cm lockchain Ar addr Show all threads a particular thread at address .Ar addr is waiting on based on non-spin locks. .\" .Pp .It Ic show Cm lockedbufs Show the same information as "show buf", but for every locked .Vt struct buf object. .\" .Pp .It Ic show Cm lockedvnods List all locked vnodes in the system. .\" .Pp .It Ic show Cm locks Prints all locks that are currently acquired. This command is only available if .Xr witness 4 is included in the kernel. .\" .Pp .It Ic show Cm locktree .\" .Pp .It Ic show Cm malloc Ns Op Li / Ns Cm i Prints .Xr malloc 9 memory allocator statistics. If the .Cm i modifier is specified, format output as machine-parseable comma-separated values ("CSV"). The output columns are as follows: .Pp .Bl -tag -compact -offset indent -width "Requests" .It Ic Type Specifies a type of memory. It is the same as a description string used while defining the given memory type with .Xr MALLOC_DECLARE 9 . .It Ic InUse Number of memory allocations of the given type, for which .Xr free 9 has not been called yet. .It Ic MemUse Total memory consumed by the given allocation type. .It Ic Requests Number of memory allocation requests for the given memory type. .El .Pp The same information can be gathered in userspace with .Dq Nm vmstat Fl m . .\" .Pp .It Ic show Cm map Ns Oo Li / Ns Cm f Oc Ar addr Prints the VM map at .Ar addr . If the .Cm f modifier is specified the complete map is printed. .\" .Pp .It Ic show Cm msgbuf Print the system's message buffer. It is the same output as in the .Dq Nm dmesg case. It is useful if you got a kernel panic, attached a serial cable to the machine and want to get the boot messages from before the system hang. .\" .Pp .It Ic show Cm mount Op Ar addr Displays details about the mount point located at .Ar addr . If no .Ar addr is specified, displays short info about all currently mounted file systems. .\" .Pp .It Ic show Cm object Ns Oo Li / Ns Cm f Oc Ar addr Prints the VM object at .Ar addr . If the .Cm f option is specified the complete object is printed. .\" .Pp .It Ic show Cm panic Print the panic message if set. .\" .Pp .It Ic show Cm page Show statistics on VM pages. .\" .Pp .It Ic show Cm pageq Show statistics on VM page queues. .\" .Pp .It Ic show Cm pciregs Print PCI bus registers. The same information can be gathered in userspace by running .Dq Nm pciconf Fl lv . .\" .Pp .It Ic show Cm pcpu Print current processor state. The output format is as follows: .Pp .Bl -tag -compact -offset indent -width "spin locks held:" .It Ic cpuid Processor identifier. .It Ic curthread Thread pointer, process identifier and the name of the process. .It Ic curpcb Control block pointer. .It Ic fpcurthread FPU thread pointer. .It Ic idlethread Idle thread pointer. .It Ic APIC ID CPU identifier coming from APIC. .It Ic currentldt LDT pointer. .It Ic spin locks held Names of spin locks held. .El .\" .Pp .It Ic show Cm pgrpdump Dump process groups present within the system. .\" .Pp .It Ic show Cm prison Op Ar addr Show the prison structure located at .Ar addr . If no .Ar addr argument is specified, show information about all prisons in the system. .\" .Pp .It Ic show Cm proc Op Ar addr Show information about the process structure located at address .Ar addr , or the current process if no argument is specified. .\" .Pp .It Ic show Cm procvm Op Ar addr Show process virtual memory layout for the process located at .Ar addr , or the current process if no argument is specified. .\" .Pp .It Ic show Cm protosw Ar addr Print protocol switch structure .Vt struct protosw at address .Ar addr . .\" .Pp .It Ic show Cm registers Ns Op Li / Ns Cm u Display the register set. If the .Cm u modifier is specified, the register contents of the thread's previous trapframe are displayed instead. Usually, this corresponds to the saved state from userspace. .\" .Pp .It Ic show Cm rman Ar addr Show resource manager object .Vt struct rman at address .Ar addr . Addresses of particular pointers can be gathered with "show allrman" command. .\" .Pp .It Ic show Cm route Ar addr Show route table result for destination .Ar addr . At this time, INET and INET6 formatted addresses are supported. .\" .Pp .It Ic show Cm routetable Oo Ar af Oc Show full route table or tables. If .Ar af is specified, show only routes for the given numeric address family. If no argument is specified, dump the route table for all address families. .\" .Pp .It Ic show Cm rtc Show real time clock value. Useful for long debugging sessions. .\" .Pp .It Ic show Cm sleepchain Deprecated. Now an alias for .Ic show Cm lockchain . .\" .Pp .It Ic show Cm sleepq Ar addr .It Ic show Cm sleepqueue Ar addr Show the .Xr sleepqueue 9 structure located at .Ar addr . .\" .Pp .It Ic show Cm sockbuf Ar addr Show the socket buffer .Va struct sockbuf located at .Ar addr . .\" .Pp .It Ic show Cm socket Ar addr Show the socket object .Vt struct socket located at .Ar addr . .\" .Pp .It Ic show Cm sysregs Show system registers (e.g., .Li cr0-4 on i386.) Not present on some platforms. .\" .Pp -.It Ic show Cm tcpcb Ns Oo Li / Ns Cm b Oc Ar addr +.It Ic show Cm tcpcb Ns Oo Li / Ns Cm b Ns Cm i Oc Ar addr Print TCP control block .Vt struct tcpcb lying at address .Ar addr . For exact interpretation of output, visit .Pa netinet/tcp.h header file. The .Cm b modifier will request BBLog entries to be printed. +If the +.Cm i +modifier is provided, the corresponding IP control block is also shown. .\" .Pp .It Ic show Cm thread Op Ar addr | tid If no .Ar addr or .Ar tid is specified, show detailed information about current thread. Otherwise, print information about the thread with ID .Ar tid or kernel address .Ar addr . (If the argument is a decimal number, it is assumed to be a tid.) .\" .Pp .It Ic show Cm threads Show all threads within the system. Output format is as follows: .Pp .Bl -tag -compact -offset indent -width "Second column" .It Ic First column Thread identifier (TID) .It Ic Second column Thread structure address .It Ic Third column Backtrace. .El .\" .Pp .It Ic show Cm tty Ar addr Display the contents of a TTY structure in a readable form. .\" .Pp .It Ic show Cm turnstile Ar addr Show turnstile .Vt struct turnstile structure at address .Ar addr . Turnstiles are structures used within the .Fx kernel to implement synchronization primitives which, while holding a specific type of lock, cannot sleep or context switch to another thread. Currently, those are: .Xr mutex 9 , .Xr rwlock 9 , .Xr rmlock 9 . .\" .Pp .It Ic show Cm uma Ns Op Li / Ns Cm i Show UMA allocator statistics. If the .Cm i modifier is specified, format output as machine-parseable comma-separated values ("CSV"). The output contains the following columns: .Pp .Bl -tag -compact -offset indent -width "Total Mem" .It Cm "Zone" Name of the UMA zone. The same string that was passed to .Xr uma_zcreate 9 as a first argument. .It Cm "Size" Size of a given memory object (slab). .It Cm "Used" Number of slabs being currently used. .It Cm "Free" Number of free slabs within the UMA zone. .It Cm "Requests" Number of allocations requests to the given zone. .It Cm "Total Mem" Total memory in use (either allocated or free) by a zone, in bytes. .It Cm "XFree" Number of free slabs within the UMA zone that were freed on a different NUMA domain than allocated. (The count in the .Cm "Free" column is inclusive of .Cm "XFree" . ) .El .Pp The same information might be gathered in the userspace with the help of .Dq Nm vmstat Fl z . .\" .Pp .It Ic show Cm unpcb Ar addr Shows UNIX domain socket private control block .Vt struct unpcb present at the address .Ar addr . .\" .Pp .It Ic show Cm vmochk Prints, whether the internal VM objects are in a map somewhere and none have zero ref counts. .\" .Pp .It Ic show Cm vmopag Walk the list of VM objects in the system, printing the indices and physical addresses of the VM pages belonging to each object. .\" .Pp .It Ic show Cm vnet Ar addr Prints virtualized network stack .Vt struct vnet structure present at the address .Ar addr . .\" .Pp .It Ic show Cm vnode Ar addr Prints vnode .Vt struct vnode structure lying at .Ar addr . For the exact interpretation of the output, look at the .Pa sys/vnode.h header file. .\" .Pp .It Ic show Cm vnodebufs Ar addr Shows clean/dirty buffer lists of the vnode located at .Ar addr . .\" .Pp .It Ic show Cm vpath Ar addr Walk the namecache to lookup the pathname of the vnode located at .Ar addr . .\" .Pp .It Ic show Cm watches Displays all watchpoints. Shows watchpoints set with "watch" command. .\" .Pp .It Ic show Cm witness Shows information about lock acquisition coming from the .Xr witness 4 subsystem. .El .Ss OFFLINE DEBUGGING COMMANDS .Bl -tag -width indent -compact .It Ic dump Initiate a kernel core dump to the device(s) configured by .Xr dumpon 8 . .Pp .It Ic gdb Switches to remote GDB mode. In remote GDB mode, another machine is required that runs .Xr gdb 1 Pq Pa ports/devel/gdb using the remote debug feature, with a connection to the serial console port on the target machine. .Pp .It Ic netdump Fl s Ar server Oo Fl g Ar gateway Fl c Ar client Fl i Ar iface Oc Configure .Xr netdump 4 with the provided parameters, and immediately perform a netdump. .Pp There are some known limitations. Principally, .Xr netdump 4 only supports IPv4 at this time. The address arguments to the .Ic netdump command must be dotted decimal IPv4 addresses. (Hostnames are not supported.) At present, the command only works if the machine is in a panic state. Finally, the .Nm .Ic netdump command does not provide any way to configure compression or encryption. .Pp .It Ic netgdb Fl s Ar server Oo Fl g Ar gateway Fl c Ar client Fl i Ar iface Oc Initiate a .Xr netgdb 4 session with the provided parameters. .Pp .Ic netgdb has identical limitations to .Ic netdump . .Pp .It Ic capture on .It Ic capture off .It Ic capture reset .It Ic capture status .Nm supports a basic output capture facility, which can be used to retrieve the results of debugging commands from userspace using .Xr sysctl 3 . .Ic capture on enables output capture; .Ic capture off disables capture. .Ic capture reset will clear the capture buffer and disable capture. .Ic capture status will report current buffer use, buffer size, and disposition of output capture. .Pp Userspace processes may inspect and manage .Nm capture state using .Xr sysctl 8 : .Pp .Va debug.ddb.capture.bufsize may be used to query or set the current capture buffer size. .Pp .Va debug.ddb.capture.maxbufsize may be used to query the compile-time limit on the capture buffer size. .Pp .Va debug.ddb.capture.bytes may be used to query the number of bytes of output currently in the capture buffer. .Pp .Va debug.ddb.capture.data returns the contents of the buffer as a string to an appropriately privileged process. .Pp This facility is particularly useful in concert with the scripting and .Xr textdump 4 facilities, allowing scripted debugging output to be captured and committed to disk as part of a textdump for later analysis. The contents of the capture buffer may also be inspected in a kernel core dump using .Xr kgdb 1 Pq Pa ports/devel/gdb . .Pp .It Ic run .It Ic script .It Ic scripts .It Ic unscript Run, define, list, and delete scripts. See the .Sx SCRIPTING section for more information on the scripting facility. .Pp .It Ic textdump dump .It Ic textdump set .It Ic textdump status .It Ic textdump unset Use the .Ic textdump dump command to immediately perform a textdump. More information may be found in .Xr textdump 4 . The .Ic textdump set command may be used to force the next kernel core dump to be a textdump rather than a traditional memory dump or minidump. .Ic textdump status reports whether a textdump has been scheduled. .Ic textdump unset cancels a request to perform a textdump as the next kernel core dump. .El .Sh VARIABLES The debugger accesses registers and variables as .Li $ Ns Ar name . Register names are as in the .Dq Ic show Cm registers command. Some variables are suffixed with numbers, and may have some modifier following a colon immediately after the variable name. For example, register variables can have a .Cm u modifier to indicate user register (e.g., .Dq Li $eax:u ) . .Pp Built-in variables currently supported are: .Pp .Bl -tag -width ".Va tabstops" -compact .It Va radix Input and output radix. .It Va maxoff Addresses are printed as .Dq Ar symbol Ns Li + Ns Ar offset unless .Ar offset is greater than .Va maxoff . .It Va maxwidth The width of the displayed line. .It Va lines The number of lines. It is used by the built-in pager. Setting it to 0 disables paging. .It Va tabstops Tab stop width. .It Va work Ns Ar xx Work variable; .Ar xx can take values from 0 to 31. .El .Sh EXPRESSIONS Most expression operators in C are supported except .Ql ~ , .Ql ^ , and unary .Ql & . Special rules in .Nm are: .Bl -tag -width ".No Identifiers" .It Identifiers The name of a symbol is translated to the value of the symbol, which is the address of the corresponding object. .Ql \&. and .Ql \&: can be used in the identifier. If supported by an object format dependent routine, .Sm off .Oo Ar filename : Oc Ar func : lineno , .Sm on .Oo Ar filename : Oc Ns Ar variable , and .Oo Ar filename : Oc Ns Ar lineno can be accepted as a symbol. .It Numbers Radix is determined by the first two letters: .Ql 0x : hex, .Ql 0o : octal, .Ql 0t : decimal; otherwise, follow current radix. .It Li \&. .Va dot .It Li + .Va next .It Li .. address of the start of the last line examined. Unlike .Va dot or .Va next , this is only changed by .Ic examine or .Ic write command. .It Li ' last address explicitly specified. .It Li $ Ns Ar variable Translated to the value of the specified variable. It may be followed by a .Ql \&: and modifiers as described above. .It Ar a Ns Li # Ns Ar b A binary operator which rounds up the left hand side to the next multiple of right hand side. .It Li * Ns Ar expr Indirection. It may be followed by a .Ql \&: and modifiers as described above. .El .Sh SCRIPTING .Nm supports a basic scripting facility to allow automating tasks or responses to specific events. Each script consists of a list of DDB commands to be executed sequentially, and is assigned a unique name. Certain script names have special meaning, and will be automatically run on various .Nm events if scripts by those names have been defined. .Pp The .Ic script command may be used to define a script by name. Scripts consist of a series of .Nm commands separated with the .Ql \&; character. For example: .Bd -literal -offset indent script kdb.enter.panic=bt; show pcpu script lockinfo=show alllocks; show lockedvnods .Ed .Pp The .Ic scripts command lists currently defined scripts. .Pp The .Ic run command execute a script by name. For example: .Bd -literal -offset indent run lockinfo .Ed .Pp The .Ic unscript command may be used to delete a script by name. For example: .Bd -literal -offset indent unscript kdb.enter.panic .Ed .Pp These functions may also be performed from userspace using the .Xr ddb 8 command. .Pp Certain scripts are run automatically, if defined, for specific .Nm events. The follow scripts are run when various events occur: .Bl -tag -width kdb.enter.powerfail .It Va kdb.enter.acpi The kernel debugger was entered as a result of an .Xr acpi 4 event. .It Va kdb.enter.bootflags The kernel debugger was entered at boot as a result of the debugger boot flag being set. .It Va kdb.enter.break The kernel debugger was entered as a result of a serial or console break. .It Va kdb.enter.cam The kernel debugger was entered as a result of a .Xr CAM 4 event. .It Va kdb.enter.mac The kernel debugger was entered as a result of an assertion failure in the .Xr mac_test 4 module of the TrustedBSD MAC Framework. .It Va kdb.enter.netgraph The kernel debugger was entered as a result of a .Xr netgraph 4 event. .It Va kdb.enter.panic .Xr panic 9 was called. .It Va kdb.enter.powerpc The kernel debugger was entered as a result of an unimplemented interrupt type on the powerpc platform. .It Va kdb.enter.sysctl The kernel debugger was entered as a result of the .Va debug.kdb.enter sysctl being set. .It Va kdb.enter.unionfs The kernel debugger was entered as a result of an assertion failure in the union file system. .It Va kdb.enter.unknown The kernel debugger was entered, but no reason has been set. .It Va kdb.enter.vfslock The kernel debugger was entered as a result of a VFS lock violation. .It Va kdb.enter.watchdog The kernel debugger was entered as a result of a watchdog firing. .It Va kdb.enter.witness The kernel debugger was entered as a result of a .Xr witness 4 violation. .El .Pp In the event that none of these scripts is found, .Nm will attempt to execute a default script: .Bl -tag -width kdb.enter.powerfail .It Va kdb.enter.default The kernel debugger was entered, but a script exactly matching the reason for entering was not defined. This can be used as a catch-all to handle cases not specifically of interest; for example, .Va kdb.enter.witness might be defined to have special handling, and .Va kdb.enter.default might be defined to simply panic and reboot. .El .Sh HINTS On machines with an ISA expansion bus, a simple NMI generation card can be constructed by connecting a push button between the A01 and B01 (CHCHK# and GND) card fingers. Momentarily shorting these two fingers together may cause the bridge chipset to generate an NMI, which causes the kernel to pass control to .Nm . Some bridge chipsets do not generate a NMI on CHCHK#, so your mileage may vary. The NMI allows one to break into the debugger on a wedged machine to diagnose problems. Other bus' bridge chipsets may be able to generate NMI using bus specific methods. There are many PCI and PCIe add-in cards which can generate NMI for debugging. Modern server systems typically use IPMI to generate signals to enter the debugger. The .Va devel/ipmitool port can be used to send the .Cd chassis power diag command which delivers an NMI to the processor. Embedded systems often use JTAG for debugging, but rarely use it in combination with .Nm . .Pp Serial consoles can break to the debugger by sending a BREAK condition on the serial line. This requires a kernel built with .Cd options BREAK_TO_DEBUGGER is specified in the kernel. Most terminal emulation programs can send a break sequence with a special key sequence or menu selection. Sending the break can be difficult or even happen spuriously in some setups. An alternative method is to build a kernel with .Cd options ALT_BREAK_TO_DEBUGGER then the sequence of CR TILDE CTRL-B enters the debugger; CR TILDE CTRL-P causes a panic; and CR TILDE CTRL-R causes an immediate reboot. In all these sequences, CR represents Carriage Return and is usually sent by pressing the Enter or Return key. TILDE is the ASCII tilde character (~). CTRL-x is Control x, sent by pressing the Control key, then x, then releasing both. .Pp The break-to-debugger behavior can be enabled by setting .Xr sysctl 8 .Va debug.kdb.break_to_debugger to 1. The alt-break-to-debugger behavior can be enabled by setting .Xr sysctl 8 .Va debug.kdb.alt_break_to_debugger to 1. The debugger can be entered by setting .Xr sysctl 8 .Va debug.kdb.enter to 1. .Pp Output can be interrupted, paused, and resumed with the control characters CTRL-C, CTRL-S, and CTRL-Q. Because these control characters are received as in-band data from the console, there is an input buffer, and once that buffer fills .Nm must either stop responding to control characters or drop additional input while continuing to search for control characters. This behavior is controlled by the tunable .Xr sysctl 8 .Va debug.ddb.prioritize_control_input , which defaults to 1. The input buffer size is 512 bytes. .Sh FILES Header files mentioned in this manual page can be found below .Pa /usr/include directory. .Pp .Bl -dash -compact .It .Pa sys/buf.h .It .Pa sys/domain.h .It .Pa netinet/in_pcb.h .It .Pa sys/socket.h .It .Pa sys/vnode.h .El .Sh SEE ALSO .Xr gdb 1 Pq Pa ports/devel/gdb , .Xr kgdb 1 Pq Pa ports/devel/gdb , .Xr acpi 4 , .Xr CAM 4 , .Xr gdb 4 , .Xr mac_ddb 4 , .Xr mac_test 4 , .Xr netgraph 4 , .Xr textdump 4 , .Xr witness 4 , .Xr ddb 8 , .Xr sysctl 8 , .Xr panic 9 .Sh HISTORY The .Nm debugger was developed for Mach, and ported to .Bx 386 0.1 . This manual page translated from .Xr man 7 macros by .An Garrett Wollman . .Pp .An Robert N. M. Watson added support for .Nm output capture, .Xr textdump 4 and scripting in .Fx 7.1 . diff --git a/sys/netinet/in_pcb.c b/sys/netinet/in_pcb.c index 69b68c77ba84..2761c6979f7d 100644 --- a/sys/netinet/in_pcb.c +++ b/sys/netinet/in_pcb.c @@ -1,3661 +1,3661 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1991, 1993, 1995 * The Regents of the University of California. * Copyright (c) 2007-2009 Robert N. M. Watson * Copyright (c) 2010-2011 Juniper Networks, Inc. * Copyright (c) 2021-2022 Gleb Smirnoff * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in_pcb.c 8.4 (Berkeley) 5/24/95 */ #include #include "opt_ddb.h" #include "opt_ipsec.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ratelimit.h" #include "opt_route.h" #include "opt_rss.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef DDB #include #endif #include #include #include #include #include #include #include #include #include #include #if defined(INET) || defined(INET6) #include #include #include #include #ifdef INET #include #include #endif #include #ifdef INET6 #include #include #include #include #endif /* INET6 */ #include #endif #include #include #define INPCBLBGROUP_SIZMIN 8 #define INPCBLBGROUP_SIZMAX 256 #define INP_FREED 0x00000200 /* Went through in_pcbfree(). */ #define INP_INLBGROUP 0x01000000 /* Inserted into inpcblbgroup. */ /* * These configure the range of local port addresses assigned to * "unspecified" outgoing connections/packets/whatever. */ VNET_DEFINE(int, ipport_lowfirstauto) = IPPORT_RESERVED - 1; /* 1023 */ VNET_DEFINE(int, ipport_lowlastauto) = IPPORT_RESERVEDSTART; /* 600 */ VNET_DEFINE(int, ipport_firstauto) = IPPORT_EPHEMERALFIRST; /* 10000 */ VNET_DEFINE(int, ipport_lastauto) = IPPORT_EPHEMERALLAST; /* 65535 */ VNET_DEFINE(int, ipport_hifirstauto) = IPPORT_HIFIRSTAUTO; /* 49152 */ VNET_DEFINE(int, ipport_hilastauto) = IPPORT_HILASTAUTO; /* 65535 */ /* * Reserved ports accessible only to root. There are significant * security considerations that must be accounted for when changing these, * but the security benefits can be great. Please be careful. */ VNET_DEFINE(int, ipport_reservedhigh) = IPPORT_RESERVED - 1; /* 1023 */ VNET_DEFINE(int, ipport_reservedlow); /* Enable random ephemeral port allocation by default. */ VNET_DEFINE(int, ipport_randomized) = 1; #ifdef INET static struct inpcb *in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, uint8_t numa_domain, int fib); #define RANGECHK(var, min, max) \ if ((var) < (min)) { (var) = (min); } \ else if ((var) > (max)) { (var) = (max); } static int sysctl_net_ipport_check(SYSCTL_HANDLER_ARGS) { int error; error = sysctl_handle_int(oidp, arg1, arg2, req); if (error == 0) { RANGECHK(V_ipport_lowfirstauto, 1, IPPORT_RESERVED - 1); RANGECHK(V_ipport_lowlastauto, 1, IPPORT_RESERVED - 1); RANGECHK(V_ipport_firstauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_lastauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_hifirstauto, IPPORT_RESERVED, IPPORT_MAX); RANGECHK(V_ipport_hilastauto, IPPORT_RESERVED, IPPORT_MAX); } return (error); } #undef RANGECHK static SYSCTL_NODE(_net_inet_ip, IPPROTO_IP, portrange, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, "IP Ports"); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowfirst, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &VNET_NAME(ipport_lowfirstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, lowlast, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &VNET_NAME(ipport_lowlastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, first, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &VNET_NAME(ipport_firstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, last, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &VNET_NAME(ipport_lastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hifirst, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &VNET_NAME(ipport_hifirstauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_PROC(_net_inet_ip_portrange, OID_AUTO, hilast, CTLFLAG_VNET | CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &VNET_NAME(ipport_hilastauto), 0, &sysctl_net_ipport_check, "I", ""); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedhigh, CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_SECURE, &VNET_NAME(ipport_reservedhigh), 0, ""); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, reservedlow, CTLFLAG_RW|CTLFLAG_SECURE, &VNET_NAME(ipport_reservedlow), 0, ""); SYSCTL_INT(_net_inet_ip_portrange, OID_AUTO, randomized, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ipport_randomized), 0, "Enable random port allocation"); #ifdef RATELIMIT counter_u64_t rate_limit_new; counter_u64_t rate_limit_chg; counter_u64_t rate_limit_active; counter_u64_t rate_limit_alloc_fail; counter_u64_t rate_limit_set_ok; static SYSCTL_NODE(_net_inet_ip, OID_AUTO, rl, CTLFLAG_RD | CTLFLAG_MPSAFE, 0, "IP Rate Limiting"); SYSCTL_COUNTER_U64(_net_inet_ip_rl, OID_AUTO, active, CTLFLAG_RD, &rate_limit_active, "Active rate limited connections"); SYSCTL_COUNTER_U64(_net_inet_ip_rl, OID_AUTO, alloc_fail, CTLFLAG_RD, &rate_limit_alloc_fail, "Rate limited connection failures"); SYSCTL_COUNTER_U64(_net_inet_ip_rl, OID_AUTO, set_ok, CTLFLAG_RD, &rate_limit_set_ok, "Rate limited setting succeeded"); SYSCTL_COUNTER_U64(_net_inet_ip_rl, OID_AUTO, newrl, CTLFLAG_RD, &rate_limit_new, "Total Rate limit new attempts"); SYSCTL_COUNTER_U64(_net_inet_ip_rl, OID_AUTO, chgrl, CTLFLAG_RD, &rate_limit_chg, "Total Rate limited change attempts"); #endif /* RATELIMIT */ #endif /* INET */ VNET_DEFINE(uint32_t, in_pcbhashseed); static void in_pcbhashseed_init(void) { V_in_pcbhashseed = arc4random(); } VNET_SYSINIT(in_pcbhashseed_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FIRST, in_pcbhashseed_init, NULL); #ifdef INET VNET_DEFINE_STATIC(int, connect_inaddr_wild) = 1; #define V_connect_inaddr_wild VNET(connect_inaddr_wild) SYSCTL_INT(_net_inet_ip, OID_AUTO, connect_inaddr_wild, CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(connect_inaddr_wild), 0, "Allow connecting to INADDR_ANY or INADDR_BROADCAST for connect(2)"); #endif static void in_pcbremhash(struct inpcb *); /* * in_pcb.c: manage the Protocol Control Blocks. * * NOTE: It is assumed that most of these functions will be called with * the pcbinfo lock held, and often, the inpcb lock held, as these utility * functions often modify hash chains or addresses in pcbs. */ static struct inpcblbgroup * in_pcblbgroup_alloc(struct ucred *cred, u_char vflag, uint16_t port, const union in_dependaddr *addr, int size, uint8_t numa_domain, int fib) { struct inpcblbgroup *grp; size_t bytes; bytes = __offsetof(struct inpcblbgroup, il_inp[size]); grp = malloc(bytes, M_PCB, M_ZERO | M_NOWAIT); if (grp == NULL) return (NULL); grp->il_cred = crhold(cred); grp->il_vflag = vflag; grp->il_lport = port; grp->il_numa_domain = numa_domain; grp->il_fibnum = fib; grp->il_dependladdr = *addr; grp->il_inpsiz = size; return (grp); } static void in_pcblbgroup_free_deferred(epoch_context_t ctx) { struct inpcblbgroup *grp; grp = __containerof(ctx, struct inpcblbgroup, il_epoch_ctx); crfree(grp->il_cred); free(grp, M_PCB); } static void in_pcblbgroup_free(struct inpcblbgroup *grp) { CK_LIST_REMOVE(grp, il_list); NET_EPOCH_CALL(in_pcblbgroup_free_deferred, &grp->il_epoch_ctx); } static void in_pcblbgroup_insert(struct inpcblbgroup *grp, struct inpcb *inp) { KASSERT(grp->il_inpcnt < grp->il_inpsiz, ("invalid local group size %d and count %d", grp->il_inpsiz, grp->il_inpcnt)); INP_WLOCK_ASSERT(inp); inp->inp_flags |= INP_INLBGROUP; grp->il_inp[grp->il_inpcnt] = inp; /* * Synchronize with in_pcblookup_lbgroup(): make sure that we don't * expose a null slot to the lookup path. */ atomic_store_rel_int(&grp->il_inpcnt, grp->il_inpcnt + 1); } static struct inpcblbgroup * in_pcblbgroup_resize(struct inpcblbgrouphead *hdr, struct inpcblbgroup *old_grp, int size) { struct inpcblbgroup *grp; int i; grp = in_pcblbgroup_alloc(old_grp->il_cred, old_grp->il_vflag, old_grp->il_lport, &old_grp->il_dependladdr, size, old_grp->il_numa_domain, old_grp->il_fibnum); if (grp == NULL) return (NULL); KASSERT(old_grp->il_inpcnt < grp->il_inpsiz, ("invalid new local group size %d and old local group count %d", grp->il_inpsiz, old_grp->il_inpcnt)); for (i = 0; i < old_grp->il_inpcnt; ++i) grp->il_inp[i] = old_grp->il_inp[i]; grp->il_inpcnt = old_grp->il_inpcnt; CK_LIST_INSERT_HEAD(hdr, grp, il_list); in_pcblbgroup_free(old_grp); return (grp); } /* * Add PCB to load balance group for SO_REUSEPORT_LB option. */ static int in_pcbinslbgrouphash(struct inpcb *inp, uint8_t numa_domain) { const static struct timeval interval = { 60, 0 }; static struct timeval lastprint; struct inpcbinfo *pcbinfo; struct inpcblbgrouphead *hdr; struct inpcblbgroup *grp; uint32_t idx; int fib; pcbinfo = inp->inp_pcbinfo; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); fib = (inp->inp_flags & INP_BOUNDFIB) != 0 ? inp->inp_inc.inc_fibnum : RT_ALL_FIBS; #ifdef INET6 /* * Don't allow IPv4 mapped INET6 wild socket. */ if ((inp->inp_vflag & INP_IPV4) && inp->inp_laddr.s_addr == INADDR_ANY && INP_CHECK_SOCKAF(inp->inp_socket, AF_INET6)) { return (0); } #endif idx = INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_lbgrouphashmask); hdr = &pcbinfo->ipi_lbgrouphashbase[idx]; CK_LIST_FOREACH(grp, hdr, il_list) { if (grp->il_cred->cr_prison == inp->inp_cred->cr_prison && grp->il_vflag == inp->inp_vflag && grp->il_lport == inp->inp_lport && grp->il_numa_domain == numa_domain && grp->il_fibnum == fib && memcmp(&grp->il_dependladdr, &inp->inp_inc.inc_ie.ie_dependladdr, sizeof(grp->il_dependladdr)) == 0) { break; } } if (grp == NULL) { /* Create new load balance group. */ grp = in_pcblbgroup_alloc(inp->inp_cred, inp->inp_vflag, inp->inp_lport, &inp->inp_inc.inc_ie.ie_dependladdr, INPCBLBGROUP_SIZMIN, numa_domain, fib); if (grp == NULL) return (ENOBUFS); in_pcblbgroup_insert(grp, inp); CK_LIST_INSERT_HEAD(hdr, grp, il_list); } else if (grp->il_inpcnt == grp->il_inpsiz) { if (grp->il_inpsiz >= INPCBLBGROUP_SIZMAX) { if (ratecheck(&lastprint, &interval)) printf("lb group port %d, limit reached\n", ntohs(grp->il_lport)); return (0); } /* Expand this local group. */ grp = in_pcblbgroup_resize(hdr, grp, grp->il_inpsiz * 2); if (grp == NULL) return (ENOBUFS); in_pcblbgroup_insert(grp, inp); } else { in_pcblbgroup_insert(grp, inp); } return (0); } /* * Remove PCB from load balance group. */ static void in_pcbremlbgrouphash(struct inpcb *inp) { struct inpcbinfo *pcbinfo; struct inpcblbgrouphead *hdr; struct inpcblbgroup *grp; int i; pcbinfo = inp->inp_pcbinfo; INP_WLOCK_ASSERT(inp); MPASS(inp->inp_flags & INP_INLBGROUP); INP_HASH_WLOCK_ASSERT(pcbinfo); hdr = &pcbinfo->ipi_lbgrouphashbase[ INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_lbgrouphashmask)]; CK_LIST_FOREACH(grp, hdr, il_list) { for (i = 0; i < grp->il_inpcnt; ++i) { if (grp->il_inp[i] != inp) continue; if (grp->il_inpcnt == 1) { /* We are the last, free this local group. */ in_pcblbgroup_free(grp); } else { KASSERT(grp->il_inpcnt >= 2, ("invalid local group count %d", grp->il_inpcnt)); grp->il_inp[i] = grp->il_inp[grp->il_inpcnt - 1]; /* * Synchronize with in_pcblookup_lbgroup(). */ atomic_store_rel_int(&grp->il_inpcnt, grp->il_inpcnt - 1); } inp->inp_flags &= ~INP_INLBGROUP; return; } } KASSERT(0, ("%s: did not find %p", __func__, inp)); } int in_pcblbgroup_numa(struct inpcb *inp, int arg) { struct inpcbinfo *pcbinfo; struct inpcblbgrouphead *hdr; struct inpcblbgroup *grp; int err, i; uint8_t numa_domain; switch (arg) { case TCP_REUSPORT_LB_NUMA_NODOM: numa_domain = M_NODOM; break; case TCP_REUSPORT_LB_NUMA_CURDOM: numa_domain = PCPU_GET(domain); break; default: if (arg < 0 || arg >= vm_ndomains) return (EINVAL); numa_domain = arg; } err = 0; pcbinfo = inp->inp_pcbinfo; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK(pcbinfo); hdr = &pcbinfo->ipi_lbgrouphashbase[ INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_lbgrouphashmask)]; CK_LIST_FOREACH(grp, hdr, il_list) { for (i = 0; i < grp->il_inpcnt; ++i) { if (grp->il_inp[i] != inp) continue; if (grp->il_numa_domain == numa_domain) { goto abort_with_hash_wlock; } /* Remove it from the old group. */ in_pcbremlbgrouphash(inp); /* Add it to the new group based on numa domain. */ in_pcbinslbgrouphash(inp, numa_domain); goto abort_with_hash_wlock; } } err = ENOENT; abort_with_hash_wlock: INP_HASH_WUNLOCK(pcbinfo); return (err); } /* Make sure it is safe to use hashinit(9) on CK_LIST. */ CTASSERT(sizeof(struct inpcbhead) == sizeof(LIST_HEAD(, inpcb))); /* * Initialize an inpcbinfo - a per-VNET instance of connections db. */ void in_pcbinfo_init(struct inpcbinfo *pcbinfo, struct inpcbstorage *pcbstor, u_int hash_nelements, u_int porthash_nelements) { mtx_init(&pcbinfo->ipi_lock, pcbstor->ips_infolock_name, NULL, MTX_DEF); mtx_init(&pcbinfo->ipi_hash_lock, pcbstor->ips_hashlock_name, NULL, MTX_DEF); #ifdef VIMAGE pcbinfo->ipi_vnet = curvnet; #endif CK_LIST_INIT(&pcbinfo->ipi_listhead); pcbinfo->ipi_count = 0; pcbinfo->ipi_hash_exact = hashinit(hash_nelements, M_PCB, &pcbinfo->ipi_hashmask); pcbinfo->ipi_hash_wild = hashinit(hash_nelements, M_PCB, &pcbinfo->ipi_hashmask); porthash_nelements = imin(porthash_nelements, IPPORT_MAX + 1); pcbinfo->ipi_porthashbase = hashinit(porthash_nelements, M_PCB, &pcbinfo->ipi_porthashmask); pcbinfo->ipi_lbgrouphashbase = hashinit(porthash_nelements, M_PCB, &pcbinfo->ipi_lbgrouphashmask); pcbinfo->ipi_zone = pcbstor->ips_zone; pcbinfo->ipi_portzone = pcbstor->ips_portzone; pcbinfo->ipi_smr = uma_zone_get_smr(pcbinfo->ipi_zone); } /* * Destroy an inpcbinfo. */ void in_pcbinfo_destroy(struct inpcbinfo *pcbinfo) { KASSERT(pcbinfo->ipi_count == 0, ("%s: ipi_count = %u", __func__, pcbinfo->ipi_count)); hashdestroy(pcbinfo->ipi_hash_exact, M_PCB, pcbinfo->ipi_hashmask); hashdestroy(pcbinfo->ipi_hash_wild, M_PCB, pcbinfo->ipi_hashmask); hashdestroy(pcbinfo->ipi_porthashbase, M_PCB, pcbinfo->ipi_porthashmask); hashdestroy(pcbinfo->ipi_lbgrouphashbase, M_PCB, pcbinfo->ipi_lbgrouphashmask); mtx_destroy(&pcbinfo->ipi_hash_lock); mtx_destroy(&pcbinfo->ipi_lock); } /* * Initialize a pcbstorage - per protocol zones to allocate inpcbs. */ static void inpcb_fini(void *, int); void in_pcbstorage_init(void *arg) { struct inpcbstorage *pcbstor = arg; pcbstor->ips_zone = uma_zcreate(pcbstor->ips_zone_name, pcbstor->ips_size, NULL, NULL, pcbstor->ips_pcbinit, inpcb_fini, UMA_ALIGN_CACHE, UMA_ZONE_SMR); pcbstor->ips_portzone = uma_zcreate(pcbstor->ips_portzone_name, sizeof(struct inpcbport), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); uma_zone_set_smr(pcbstor->ips_portzone, uma_zone_get_smr(pcbstor->ips_zone)); } /* * Destroy a pcbstorage - used by unloadable protocols. */ void in_pcbstorage_destroy(void *arg) { struct inpcbstorage *pcbstor = arg; uma_zdestroy(pcbstor->ips_zone); uma_zdestroy(pcbstor->ips_portzone); } /* * Allocate a PCB and associate it with the socket. * On success return with the PCB locked. */ int in_pcballoc(struct socket *so, struct inpcbinfo *pcbinfo) { struct inpcb *inp; #if defined(IPSEC) || defined(IPSEC_SUPPORT) || defined(MAC) int error; #endif inp = uma_zalloc_smr(pcbinfo->ipi_zone, M_NOWAIT); if (inp == NULL) return (ENOBUFS); bzero(&inp->inp_start_zero, inp_zero_size); #ifdef NUMA inp->inp_numa_domain = M_NODOM; #endif inp->inp_pcbinfo = pcbinfo; inp->inp_socket = so; inp->inp_cred = crhold(so->so_cred); inp->inp_inc.inc_fibnum = so->so_fibnum; #ifdef MAC error = mac_inpcb_init(inp, M_NOWAIT); if (error != 0) goto out; mac_inpcb_create(so, inp); #endif #if defined(IPSEC) || defined(IPSEC_SUPPORT) error = ipsec_init_pcbpolicy(inp); if (error != 0) { #ifdef MAC mac_inpcb_destroy(inp); #endif goto out; } #endif /*IPSEC*/ #ifdef INET6 if (INP_SOCKAF(so) == AF_INET6) { inp->inp_vflag |= INP_IPV6PROTO | INP_IPV6; if (V_ip6_v6only) inp->inp_flags |= IN6P_IPV6_V6ONLY; #ifdef INET else inp->inp_vflag |= INP_IPV4; #endif if (V_ip6_auto_flowlabel) inp->inp_flags |= IN6P_AUTOFLOWLABEL; inp->in6p_hops = -1; /* use kernel default */ } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET inp->inp_vflag |= INP_IPV4; #endif inp->inp_smr = SMR_SEQ_INVALID; /* * Routes in inpcb's can cache L2 as well; they are guaranteed * to be cleaned up. */ inp->inp_route.ro_flags = RT_LLE_CACHE; refcount_init(&inp->inp_refcount, 1); /* Reference from socket. */ INP_WLOCK(inp); INP_INFO_WLOCK(pcbinfo); pcbinfo->ipi_count++; inp->inp_gencnt = ++pcbinfo->ipi_gencnt; CK_LIST_INSERT_HEAD(&pcbinfo->ipi_listhead, inp, inp_list); INP_INFO_WUNLOCK(pcbinfo); so->so_pcb = inp; return (0); #if defined(IPSEC) || defined(IPSEC_SUPPORT) || defined(MAC) out: crfree(inp->inp_cred); #ifdef INVARIANTS inp->inp_cred = NULL; #endif uma_zfree_smr(pcbinfo->ipi_zone, inp); return (error); #endif } #ifdef INET int in_pcbbind(struct inpcb *inp, struct sockaddr_in *sin, int flags, struct ucred *cred) { int anonport, error; KASSERT(sin == NULL || sin->sin_family == AF_INET, ("%s: invalid address family for %p", __func__, sin)); KASSERT(sin == NULL || sin->sin_len == sizeof(struct sockaddr_in), ("%s: invalid address length for %p", __func__, sin)); INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); if (inp->inp_lport != 0 || inp->inp_laddr.s_addr != INADDR_ANY) return (EINVAL); anonport = sin == NULL || sin->sin_port == 0; error = in_pcbbind_setup(inp, sin, &inp->inp_laddr.s_addr, &inp->inp_lport, flags, cred); if (error) return (error); if (in_pcbinshash(inp) != 0) { inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_lport = 0; inp->inp_flags &= ~INP_BOUNDFIB; return (EAGAIN); } if (anonport) inp->inp_flags |= INP_ANONPORT; return (0); } #endif #if defined(INET) || defined(INET6) /* * Assign a local port like in_pcb_lport(), but also used with connect() * and a foreign address and port. If fsa is non-NULL, choose a local port * that is unused with those, otherwise one that is completely unused. * lsa can be NULL for IPv6. */ int in_pcb_lport_dest(struct inpcb *inp, struct sockaddr *lsa, u_short *lportp, struct sockaddr *fsa, u_short fport, struct ucred *cred, int lookupflags) { struct inpcbinfo *pcbinfo; struct inpcb *tmpinp; unsigned short *lastport; int count, error; u_short aux, first, last, lport; #ifdef INET struct in_addr laddr, faddr; #endif #ifdef INET6 struct in6_addr *laddr6, *faddr6; #endif pcbinfo = inp->inp_pcbinfo; /* * Because no actual state changes occur here, a global write lock on * the pcbinfo isn't required. */ INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(pcbinfo); if (inp->inp_flags & INP_HIGHPORT) { first = V_ipport_hifirstauto; /* sysctl */ last = V_ipport_hilastauto; lastport = &pcbinfo->ipi_lasthi; } else if (inp->inp_flags & INP_LOWPORT) { error = priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT); if (error) return (error); first = V_ipport_lowfirstauto; /* 1023 */ last = V_ipport_lowlastauto; /* 600 */ lastport = &pcbinfo->ipi_lastlow; } else { first = V_ipport_firstauto; /* sysctl */ last = V_ipport_lastauto; lastport = &pcbinfo->ipi_lastport; } /* * Instead of having two loops further down counting up or down * make sure that first is always <= last and go with only one * code path implementing all logic. */ if (first > last) { aux = first; first = last; last = aux; } #ifdef INET laddr.s_addr = INADDR_ANY; /* used by INET6+INET below too */ if ((inp->inp_vflag & (INP_IPV4|INP_IPV6)) == INP_IPV4) { if (lsa != NULL) laddr = ((struct sockaddr_in *)lsa)->sin_addr; if (fsa != NULL) faddr = ((struct sockaddr_in *)fsa)->sin_addr; } #endif #ifdef INET6 laddr6 = NULL; if ((inp->inp_vflag & INP_IPV6) != 0) { if (lsa != NULL) laddr6 = &((struct sockaddr_in6 *)lsa)->sin6_addr; if (fsa != NULL) faddr6 = &((struct sockaddr_in6 *)fsa)->sin6_addr; } #endif tmpinp = NULL; lport = *lportp; if (V_ipport_randomized) *lastport = first + (arc4random() % (last - first)); count = last - first; do { if (count-- < 0) /* completely used? */ return (EADDRNOTAVAIL); ++*lastport; if (*lastport < first || *lastport > last) *lastport = first; lport = htons(*lastport); if (fsa != NULL) { #ifdef INET if (lsa->sa_family == AF_INET) { tmpinp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, lookupflags, M_NODOM, RT_ALL_FIBS); } #endif #ifdef INET6 if (lsa->sa_family == AF_INET6) { tmpinp = in6_pcblookup_hash_locked(pcbinfo, faddr6, fport, laddr6, lport, lookupflags, M_NODOM, RT_ALL_FIBS); } #endif } else { #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) { tmpinp = in6_pcblookup_local(pcbinfo, &inp->in6p_laddr, lport, RT_ALL_FIBS, lookupflags, cred); #ifdef INET if (tmpinp == NULL && (inp->inp_vflag & INP_IPV4)) tmpinp = in_pcblookup_local(pcbinfo, laddr, lport, RT_ALL_FIBS, lookupflags, cred); #endif } #endif #if defined(INET) && defined(INET6) else #endif #ifdef INET tmpinp = in_pcblookup_local(pcbinfo, laddr, lport, RT_ALL_FIBS, lookupflags, cred); #endif } } while (tmpinp != NULL); *lportp = lport; return (0); } /* * Select a local port (number) to use. */ int in_pcb_lport(struct inpcb *inp, struct in_addr *laddrp, u_short *lportp, struct ucred *cred, int lookupflags) { struct sockaddr_in laddr; if (laddrp) { bzero(&laddr, sizeof(laddr)); laddr.sin_family = AF_INET; laddr.sin_addr = *laddrp; } return (in_pcb_lport_dest(inp, laddrp ? (struct sockaddr *) &laddr : NULL, lportp, NULL, 0, cred, lookupflags)); } #endif /* INET || INET6 */ #ifdef INET /* * Determine whether the inpcb can be bound to the specified address/port tuple. */ static int in_pcbbind_avail(struct inpcb *inp, const struct in_addr laddr, const u_short lport, const int fib, int sooptions, int lookupflags, struct ucred *cred) { int reuseport, reuseport_lb; INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo); reuseport = (sooptions & SO_REUSEPORT); reuseport_lb = (sooptions & SO_REUSEPORT_LB); if (IN_MULTICAST(ntohl(laddr.s_addr))) { /* * Treat SO_REUSEADDR as SO_REUSEPORT for multicast; * allow complete duplication of binding if * SO_REUSEPORT is set, or if SO_REUSEADDR is set * and a multicast address is bound on both * new and duplicated sockets. */ if ((sooptions & (SO_REUSEADDR | SO_REUSEPORT)) != 0) reuseport = SO_REUSEADDR | SO_REUSEPORT; /* * XXX: How to deal with SO_REUSEPORT_LB here? * Treat same as SO_REUSEPORT for now. */ if ((sooptions & (SO_REUSEADDR | SO_REUSEPORT_LB)) != 0) reuseport_lb = SO_REUSEADDR | SO_REUSEPORT_LB; } else if (!in_nullhost(laddr)) { struct sockaddr_in sin; memset(&sin, 0, sizeof(sin)); sin.sin_family = AF_INET; sin.sin_len = sizeof(sin); sin.sin_addr = laddr; /* * Is the address a local IP address? * If INP_BINDANY is set, then the socket may be bound * to any endpoint address, local or not. */ if ((inp->inp_flags & INP_BINDANY) == 0 && ifa_ifwithaddr_check((const struct sockaddr *)&sin) == 0) return (EADDRNOTAVAIL); } if (lport != 0) { struct inpcb *t; if (ntohs(lport) <= V_ipport_reservedhigh && ntohs(lport) >= V_ipport_reservedlow && priv_check_cred(cred, PRIV_NETINET_RESERVEDPORT)) return (EACCES); if (!IN_MULTICAST(ntohl(laddr.s_addr)) && priv_check_cred(inp->inp_cred, PRIV_NETINET_REUSEPORT) != 0) { /* * If a socket owned by a different user is already * bound to this port, fail. In particular, SO_REUSE* * can only be used to share a port among sockets owned * by the same user. * * However, we can share a port with a connected socket * which has a unique 4-tuple. */ t = in_pcblookup_local(inp->inp_pcbinfo, laddr, lport, RT_ALL_FIBS, INPLOOKUP_WILDCARD, cred); if (t != NULL && (inp->inp_socket->so_type != SOCK_STREAM || in_nullhost(t->inp_faddr)) && (inp->inp_cred->cr_uid != t->inp_cred->cr_uid)) return (EADDRINUSE); } t = in_pcblookup_local(inp->inp_pcbinfo, laddr, lport, fib, lookupflags, cred); if (t != NULL && ((reuseport | reuseport_lb) & t->inp_socket->so_options) == 0) { #ifdef INET6 if (!in_nullhost(laddr) || !in_nullhost(t->inp_laddr) || (inp->inp_vflag & INP_IPV6PROTO) == 0 || (t->inp_vflag & INP_IPV6PROTO) == 0) #endif return (EADDRINUSE); } } return (0); } /* * Set up a bind operation on a PCB, performing port allocation * as required, but do not actually modify the PCB. Callers can * either complete the bind by setting inp_laddr/inp_lport and * calling in_pcbinshash(), or they can just use the resulting * port and address to authorise the sending of a once-off packet. * * On error, the values of *laddrp and *lportp are not changed. */ int in_pcbbind_setup(struct inpcb *inp, struct sockaddr_in *sin, in_addr_t *laddrp, u_short *lportp, int flags, struct ucred *cred) { struct socket *so = inp->inp_socket; struct in_addr laddr; u_short lport = 0; int error, fib, lookupflags, sooptions; /* * No state changes, so read locks are sufficient here. */ INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo); laddr.s_addr = *laddrp; if (sin != NULL && laddr.s_addr != INADDR_ANY) return (EINVAL); lookupflags = 0; sooptions = atomic_load_int(&so->so_options); if ((sooptions & (SO_REUSEADDR | SO_REUSEPORT | SO_REUSEPORT_LB)) == 0) lookupflags = INPLOOKUP_WILDCARD; if (sin == NULL) { if ((error = prison_local_ip4(cred, &laddr)) != 0) return (error); } else { KASSERT(sin->sin_family == AF_INET, ("%s: invalid family for address %p", __func__, sin)); KASSERT(sin->sin_len == sizeof(*sin), ("%s: invalid length for address %p", __func__, sin)); error = prison_local_ip4(cred, &sin->sin_addr); if (error) return (error); if (sin->sin_port != *lportp) { /* Don't allow the port to change. */ if (*lportp != 0) return (EINVAL); lport = sin->sin_port; } laddr = sin->sin_addr; fib = (flags & INPBIND_FIB) != 0 ? inp->inp_inc.inc_fibnum : RT_ALL_FIBS; /* See if this address/port combo is available. */ error = in_pcbbind_avail(inp, laddr, lport, fib, sooptions, lookupflags, cred); if (error != 0) return (error); } if (*lportp != 0) lport = *lportp; if (lport == 0) { error = in_pcb_lport(inp, &laddr, &lport, cred, lookupflags); if (error != 0) return (error); } *laddrp = laddr.s_addr; *lportp = lport; if ((flags & INPBIND_FIB) != 0) inp->inp_flags |= INP_BOUNDFIB; return (0); } /* * Connect from a socket to a specified address. * Both address and port must be specified in argument sin. * If don't have a local address for this socket yet, * then pick one. */ int in_pcbconnect(struct inpcb *inp, struct sockaddr_in *sin, struct ucred *cred, bool rehash __unused) { u_short lport, fport; in_addr_t laddr, faddr; int anonport, error; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); KASSERT(in_nullhost(inp->inp_faddr), ("%s: inp is already connected", __func__)); lport = inp->inp_lport; laddr = inp->inp_laddr.s_addr; anonport = (lport == 0); error = in_pcbconnect_setup(inp, sin, &laddr, &lport, &faddr, &fport, cred); if (error) return (error); inp->inp_faddr.s_addr = faddr; inp->inp_fport = fport; /* Do the initial binding of the local address if required. */ if (inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0) { inp->inp_lport = lport; inp->inp_laddr.s_addr = laddr; if (in_pcbinshash(inp) != 0) { inp->inp_laddr.s_addr = inp->inp_faddr.s_addr = INADDR_ANY; inp->inp_lport = inp->inp_fport = 0; return (EAGAIN); } } else { inp->inp_lport = lport; inp->inp_laddr.s_addr = laddr; if ((inp->inp_flags & INP_INHASHLIST) != 0) in_pcbrehash(inp); else in_pcbinshash(inp); } if (anonport) inp->inp_flags |= INP_ANONPORT; return (0); } /* * Do proper source address selection on an unbound socket in case * of connect. Take jails into account as well. */ int in_pcbladdr(struct inpcb *inp, struct in_addr *faddr, struct in_addr *laddr, struct ucred *cred) { struct ifaddr *ifa; struct sockaddr *sa; struct sockaddr_in *sin, dst; struct nhop_object *nh; int error; NET_EPOCH_ASSERT(); KASSERT(laddr != NULL, ("%s: laddr NULL", __func__)); /* * Bypass source address selection and use the primary jail IP * if requested. */ if (!prison_saddrsel_ip4(cred, laddr)) return (0); error = 0; nh = NULL; bzero(&dst, sizeof(dst)); sin = &dst; sin->sin_family = AF_INET; sin->sin_len = sizeof(struct sockaddr_in); sin->sin_addr.s_addr = faddr->s_addr; /* * If route is known our src addr is taken from the i/f, * else punt. * * Find out route to destination. */ if ((inp->inp_socket->so_options & SO_DONTROUTE) == 0) nh = fib4_lookup(inp->inp_inc.inc_fibnum, *faddr, 0, NHR_NONE, 0); /* * If we found a route, use the address corresponding to * the outgoing interface. * * Otherwise assume faddr is reachable on a directly connected * network and try to find a corresponding interface to take * the source address from. */ if (nh == NULL || nh->nh_ifp == NULL) { struct in_ifaddr *ia; struct ifnet *ifp; ia = ifatoia(ifa_ifwithdstaddr((struct sockaddr *)sin, inp->inp_socket->so_fibnum)); if (ia == NULL) { ia = ifatoia(ifa_ifwithnet((struct sockaddr *)sin, 0, inp->inp_socket->so_fibnum)); } if (ia == NULL) { error = ENETUNREACH; goto done; } if (!prison_flag(cred, PR_IP4)) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } ifp = ia->ia_ifp; ia = NULL; CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } /* * If the outgoing interface on the route found is not * a loopback interface, use the address from that interface. * In case of jails do those three steps: * 1. check if the interface address belongs to the jail. If so use it. * 2. check if we have any address on the outgoing interface * belonging to this jail. If so use it. * 3. as a last resort return the 'default' jail address. */ if ((nh->nh_ifp->if_flags & IFF_LOOPBACK) == 0) { struct in_ifaddr *ia; struct ifnet *ifp; /* If not jailed, use the default returned. */ if (!prison_flag(cred, PR_IP4)) { ia = (struct in_ifaddr *)nh->nh_ifa; laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* Jailed. */ /* 1. Check if the iface address belongs to the jail. */ sin = (struct sockaddr_in *)nh->nh_ifa->ifa_addr; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)nh->nh_ifa; laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* * 2. Check if we have any address on the outgoing interface * belonging to this jail. */ ia = NULL; ifp = nh->nh_ifp; CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } /* * The outgoing interface is marked with 'loopback net', so a route * to ourselves is here. * Try to find the interface of the destination address and then * take the address from there. That interface is not necessarily * a loopback interface. * In case of jails, check that it is an address of the jail * and if we cannot find, fall back to the 'default' jail address. */ if ((nh->nh_ifp->if_flags & IFF_LOOPBACK) != 0) { struct in_ifaddr *ia; ia = ifatoia(ifa_ifwithdstaddr(sintosa(&dst), inp->inp_socket->so_fibnum)); if (ia == NULL) ia = ifatoia(ifa_ifwithnet(sintosa(&dst), 0, inp->inp_socket->so_fibnum)); if (ia == NULL) ia = ifatoia(ifa_ifwithaddr(sintosa(&dst))); if (!prison_flag(cred, PR_IP4)) { if (ia == NULL) { error = ENETUNREACH; goto done; } laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } /* Jailed. */ if (ia != NULL) { struct ifnet *ifp; ifp = ia->ia_ifp; ia = NULL; CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { sa = ifa->ifa_addr; if (sa->sa_family != AF_INET) continue; sin = (struct sockaddr_in *)sa; if (prison_check_ip4(cred, &sin->sin_addr) == 0) { ia = (struct in_ifaddr *)ifa; break; } } if (ia != NULL) { laddr->s_addr = ia->ia_addr.sin_addr.s_addr; goto done; } } /* 3. As a last resort return the 'default' jail address. */ error = prison_get_ip4(cred, laddr); goto done; } done: if (error == 0 && laddr->s_addr == INADDR_ANY) return (EHOSTUNREACH); return (error); } /* * Set up for a connect from a socket to the specified address. * On entry, *laddrp and *lportp should contain the current local * address and port for the PCB; these are updated to the values * that should be placed in inp_laddr and inp_lport to complete * the connect. * * On success, *faddrp and *fportp will be set to the remote address * and port. These are not updated in the error case. */ int in_pcbconnect_setup(struct inpcb *inp, struct sockaddr_in *sin, in_addr_t *laddrp, u_short *lportp, in_addr_t *faddrp, u_short *fportp, struct ucred *cred) { struct in_ifaddr *ia; struct in_addr laddr, faddr; u_short lport, fport; int error; KASSERT(sin->sin_family == AF_INET, ("%s: invalid address family for %p", __func__, sin)); KASSERT(sin->sin_len == sizeof(*sin), ("%s: invalid address length for %p", __func__, sin)); /* * Because a global state change doesn't actually occur here, a read * lock is sufficient. */ NET_EPOCH_ASSERT(); INP_LOCK_ASSERT(inp); INP_HASH_LOCK_ASSERT(inp->inp_pcbinfo); if (sin->sin_port == 0) return (EADDRNOTAVAIL); laddr.s_addr = *laddrp; lport = *lportp; faddr = sin->sin_addr; fport = sin->sin_port; #ifdef ROUTE_MPATH if (CALC_FLOWID_OUTBOUND) { uint32_t hash_val, hash_type; hash_val = fib4_calc_software_hash(laddr, faddr, 0, fport, inp->inp_socket->so_proto->pr_protocol, &hash_type); inp->inp_flowid = hash_val; inp->inp_flowtype = hash_type; } #endif if (V_connect_inaddr_wild && !CK_STAILQ_EMPTY(&V_in_ifaddrhead)) { /* * If the destination address is INADDR_ANY, * use the primary local address. * If the supplied address is INADDR_BROADCAST, * and the primary interface supports broadcast, * choose the broadcast address for that interface. */ if (faddr.s_addr == INADDR_ANY) { faddr = IA_SIN(CK_STAILQ_FIRST(&V_in_ifaddrhead))->sin_addr; if ((error = prison_get_ip4(cred, &faddr)) != 0) return (error); } else if (faddr.s_addr == (u_long)INADDR_BROADCAST) { if (CK_STAILQ_FIRST(&V_in_ifaddrhead)->ia_ifp->if_flags & IFF_BROADCAST) faddr = satosin(&CK_STAILQ_FIRST( &V_in_ifaddrhead)->ia_broadaddr)->sin_addr; } } else if (faddr.s_addr == INADDR_ANY) { return (ENETUNREACH); } if (laddr.s_addr == INADDR_ANY) { error = in_pcbladdr(inp, &faddr, &laddr, cred); /* * If the destination address is multicast and an outgoing * interface has been set as a multicast option, prefer the * address of that interface as our source address. */ if (IN_MULTICAST(ntohl(faddr.s_addr)) && inp->inp_moptions != NULL) { struct ip_moptions *imo; struct ifnet *ifp; imo = inp->inp_moptions; if (imo->imo_multicast_ifp != NULL) { ifp = imo->imo_multicast_ifp; CK_STAILQ_FOREACH(ia, &V_in_ifaddrhead, ia_link) { if (ia->ia_ifp == ifp && prison_check_ip4(cred, &ia->ia_addr.sin_addr) == 0) break; } if (ia == NULL) error = EADDRNOTAVAIL; else { laddr = ia->ia_addr.sin_addr; error = 0; } } } if (error) return (error); } if (lport != 0) { if (in_pcblookup_hash_locked(inp->inp_pcbinfo, faddr, fport, laddr, lport, 0, M_NODOM, RT_ALL_FIBS) != NULL) return (EADDRINUSE); } else { struct sockaddr_in lsin, fsin; bzero(&lsin, sizeof(lsin)); bzero(&fsin, sizeof(fsin)); lsin.sin_family = AF_INET; lsin.sin_addr = laddr; fsin.sin_family = AF_INET; fsin.sin_addr = faddr; error = in_pcb_lport_dest(inp, (struct sockaddr *) &lsin, &lport, (struct sockaddr *)& fsin, fport, cred, INPLOOKUP_WILDCARD); if (error) return (error); } *laddrp = laddr.s_addr; *lportp = lport; *faddrp = faddr.s_addr; *fportp = fport; return (0); } void in_pcbdisconnect(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); KASSERT(inp->inp_smr == SMR_SEQ_INVALID, ("%s: inp %p was already disconnected", __func__, inp)); in_pcbremhash_locked(inp); /* See the comment in in_pcbinshash(). */ inp->inp_smr = smr_advance(inp->inp_pcbinfo->ipi_smr); inp->inp_laddr.s_addr = INADDR_ANY; inp->inp_faddr.s_addr = INADDR_ANY; inp->inp_fport = 0; } #endif /* INET */ /* * inpcb hash lookups are protected by SMR section. * * Once desired pcb has been found, switching from SMR section to a pcb * lock is performed with inp_smr_lock(). We can not use INP_(W|R)LOCK * here because SMR is a critical section. * In 99%+ cases inp_smr_lock() would obtain the lock immediately. */ void inp_lock(struct inpcb *inp, const inp_lookup_t lock) { lock == INPLOOKUP_RLOCKPCB ? rw_rlock(&inp->inp_lock) : rw_wlock(&inp->inp_lock); } void inp_unlock(struct inpcb *inp, const inp_lookup_t lock) { lock == INPLOOKUP_RLOCKPCB ? rw_runlock(&inp->inp_lock) : rw_wunlock(&inp->inp_lock); } int inp_trylock(struct inpcb *inp, const inp_lookup_t lock) { return (lock == INPLOOKUP_RLOCKPCB ? rw_try_rlock(&inp->inp_lock) : rw_try_wlock(&inp->inp_lock)); } static inline bool _inp_smr_lock(struct inpcb *inp, const inp_lookup_t lock, const int ignflags) { MPASS(lock == INPLOOKUP_RLOCKPCB || lock == INPLOOKUP_WLOCKPCB); SMR_ASSERT_ENTERED(inp->inp_pcbinfo->ipi_smr); if (__predict_true(inp_trylock(inp, lock))) { if (__predict_false(inp->inp_flags & ignflags)) { smr_exit(inp->inp_pcbinfo->ipi_smr); inp_unlock(inp, lock); return (false); } smr_exit(inp->inp_pcbinfo->ipi_smr); return (true); } if (__predict_true(refcount_acquire_if_not_zero(&inp->inp_refcount))) { smr_exit(inp->inp_pcbinfo->ipi_smr); inp_lock(inp, lock); if (__predict_false(in_pcbrele(inp, lock))) return (false); /* * inp acquired through refcount & lock for sure didn't went * through uma_zfree(). However, it may have already went * through in_pcbfree() and has another reference, that * prevented its release by our in_pcbrele(). */ if (__predict_false(inp->inp_flags & ignflags)) { inp_unlock(inp, lock); return (false); } return (true); } else { smr_exit(inp->inp_pcbinfo->ipi_smr); return (false); } } bool inp_smr_lock(struct inpcb *inp, const inp_lookup_t lock) { /* * in_pcblookup() family of functions ignore not only freed entries, * that may be found due to lockless access to the hash, but dropped * entries, too. */ return (_inp_smr_lock(inp, lock, INP_FREED | INP_DROPPED)); } /* * inp_next() - inpcb hash/list traversal iterator * * Requires initialized struct inpcb_iterator for context. * The structure can be initialized with INP_ITERATOR() or INP_ALL_ITERATOR(). * * - Iterator can have either write-lock or read-lock semantics, that can not * be changed later. * - Iterator can iterate either over all pcbs list (INP_ALL_LIST), or through * a single hash slot. Note: only rip_input() does the latter. * - Iterator may have optional bool matching function. The matching function * will be executed for each inpcb in the SMR context, so it can not acquire * locks and can safely access only immutable fields of inpcb. * * A fresh initialized iterator has NULL inpcb in its context and that * means that inp_next() call would return the very first inpcb on the list * locked with desired semantic. In all following calls the context pointer * shall hold the current inpcb pointer. The KPI user is not supposed to * unlock the current inpcb! Upon end of traversal inp_next() will return NULL * and write NULL to its context. After end of traversal an iterator can be * reused. * * List traversals have the following features/constraints: * - New entries won't be seen, as they are always added to the head of a list. * - Removed entries won't stop traversal as long as they are not added to * a different list. This is violated by in_pcbrehash(). */ #define II_LIST_FIRST(ipi, hash) \ (((hash) == INP_ALL_LIST) ? \ CK_LIST_FIRST(&(ipi)->ipi_listhead) : \ CK_LIST_FIRST(&(ipi)->ipi_hash_exact[(hash)])) #define II_LIST_NEXT(inp, hash) \ (((hash) == INP_ALL_LIST) ? \ CK_LIST_NEXT((inp), inp_list) : \ CK_LIST_NEXT((inp), inp_hash_exact)) #define II_LOCK_ASSERT(inp, lock) \ rw_assert(&(inp)->inp_lock, \ (lock) == INPLOOKUP_RLOCKPCB ? RA_RLOCKED : RA_WLOCKED ) struct inpcb * inp_next(struct inpcb_iterator *ii) { const struct inpcbinfo *ipi = ii->ipi; inp_match_t *match = ii->match; void *ctx = ii->ctx; inp_lookup_t lock = ii->lock; int hash = ii->hash; struct inpcb *inp; if (ii->inp == NULL) { /* First call. */ smr_enter(ipi->ipi_smr); /* This is unrolled CK_LIST_FOREACH(). */ for (inp = II_LIST_FIRST(ipi, hash); inp != NULL; inp = II_LIST_NEXT(inp, hash)) { if (match != NULL && (match)(inp, ctx) == false) continue; if (__predict_true(_inp_smr_lock(inp, lock, INP_FREED))) break; else { smr_enter(ipi->ipi_smr); MPASS(inp != II_LIST_FIRST(ipi, hash)); inp = II_LIST_FIRST(ipi, hash); if (inp == NULL) break; } } if (inp == NULL) smr_exit(ipi->ipi_smr); else ii->inp = inp; return (inp); } /* Not a first call. */ smr_enter(ipi->ipi_smr); restart: inp = ii->inp; II_LOCK_ASSERT(inp, lock); next: inp = II_LIST_NEXT(inp, hash); if (inp == NULL) { smr_exit(ipi->ipi_smr); goto found; } if (match != NULL && (match)(inp, ctx) == false) goto next; if (__predict_true(inp_trylock(inp, lock))) { if (__predict_false(inp->inp_flags & INP_FREED)) { /* * Entries are never inserted in middle of a list, thus * as long as we are in SMR, we can continue traversal. * Jump to 'restart' should yield in the same result, * but could produce unnecessary looping. Could this * looping be unbound? */ inp_unlock(inp, lock); goto next; } else { smr_exit(ipi->ipi_smr); goto found; } } /* * Can't obtain lock immediately, thus going hard. Once we exit the * SMR section we can no longer jump to 'next', and our only stable * anchoring point is ii->inp, which we keep locked for this case, so * we jump to 'restart'. */ if (__predict_true(refcount_acquire_if_not_zero(&inp->inp_refcount))) { smr_exit(ipi->ipi_smr); inp_lock(inp, lock); if (__predict_false(in_pcbrele(inp, lock))) { smr_enter(ipi->ipi_smr); goto restart; } /* * See comment in inp_smr_lock(). */ if (__predict_false(inp->inp_flags & INP_FREED)) { inp_unlock(inp, lock); smr_enter(ipi->ipi_smr); goto restart; } } else goto next; found: inp_unlock(ii->inp, lock); ii->inp = inp; return (ii->inp); } /* * in_pcbref() bumps the reference count on an inpcb in order to maintain * stability of an inpcb pointer despite the inpcb lock being released or * SMR section exited. * * To free a reference later in_pcbrele_(r|w)locked() must be performed. */ void in_pcbref(struct inpcb *inp) { u_int old __diagused; old = refcount_acquire(&inp->inp_refcount); KASSERT(old > 0, ("%s: refcount 0", __func__)); } /* * Drop a refcount on an inpcb elevated using in_pcbref(), potentially * freeing the pcb, if the reference was very last. */ bool in_pcbrele_rlocked(struct inpcb *inp) { INP_RLOCK_ASSERT(inp); if (!refcount_release(&inp->inp_refcount)) return (false); MPASS(inp->inp_flags & INP_FREED); MPASS(inp->inp_socket == NULL); crfree(inp->inp_cred); #ifdef INVARIANTS inp->inp_cred = NULL; #endif INP_RUNLOCK(inp); uma_zfree_smr(inp->inp_pcbinfo->ipi_zone, inp); return (true); } bool in_pcbrele_wlocked(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); if (!refcount_release(&inp->inp_refcount)) return (false); MPASS(inp->inp_flags & INP_FREED); MPASS(inp->inp_socket == NULL); crfree(inp->inp_cred); #ifdef INVARIANTS inp->inp_cred = NULL; #endif INP_WUNLOCK(inp); uma_zfree_smr(inp->inp_pcbinfo->ipi_zone, inp); return (true); } bool in_pcbrele(struct inpcb *inp, const inp_lookup_t lock) { return (lock == INPLOOKUP_RLOCKPCB ? in_pcbrele_rlocked(inp) : in_pcbrele_wlocked(inp)); } /* * Unconditionally schedule an inpcb to be freed by decrementing its * reference count, which should occur only after the inpcb has been detached * from its socket. If another thread holds a temporary reference (acquired * using in_pcbref()) then the free is deferred until that reference is * released using in_pcbrele_(r|w)locked(), but the inpcb is still unlocked. * Almost all work, including removal from global lists, is done in this * context, where the pcbinfo lock is held. */ void in_pcbfree(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; #ifdef INET struct ip_moptions *imo; #endif #ifdef INET6 struct ip6_moptions *im6o; #endif INP_WLOCK_ASSERT(inp); KASSERT(inp->inp_socket != NULL, ("%s: inp_socket == NULL", __func__)); KASSERT((inp->inp_flags & INP_FREED) == 0, ("%s: called twice for pcb %p", __func__, inp)); /* * in_pcblookup_local() and in6_pcblookup_local() may return an inpcb * from the hash without acquiring inpcb lock, they rely on the hash * lock, thus in_pcbremhash() should be the first action. */ if (inp->inp_flags & INP_INHASHLIST) in_pcbremhash(inp); INP_INFO_WLOCK(pcbinfo); inp->inp_gencnt = ++pcbinfo->ipi_gencnt; pcbinfo->ipi_count--; CK_LIST_REMOVE(inp, inp_list); INP_INFO_WUNLOCK(pcbinfo); #ifdef RATELIMIT if (inp->inp_snd_tag != NULL) in_pcbdetach_txrtlmt(inp); #endif inp->inp_flags |= INP_FREED; inp->inp_socket->so_pcb = NULL; inp->inp_socket = NULL; RO_INVALIDATE_CACHE(&inp->inp_route); #ifdef MAC mac_inpcb_destroy(inp); #endif #if defined(IPSEC) || defined(IPSEC_SUPPORT) if (inp->inp_sp != NULL) ipsec_delete_pcbpolicy(inp); #endif #ifdef INET if (inp->inp_options) (void)m_free(inp->inp_options); imo = inp->inp_moptions; #endif #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) { ip6_freepcbopts(inp->in6p_outputopts); im6o = inp->in6p_moptions; } else im6o = NULL; #endif if (__predict_false(in_pcbrele_wlocked(inp) == false)) { INP_WUNLOCK(inp); } #ifdef INET6 ip6_freemoptions(im6o); #endif #ifdef INET inp_freemoptions(imo); #endif } /* * Different protocols initialize their inpcbs differently - giving * different name to the lock. But they all are disposed the same. */ static void inpcb_fini(void *mem, int size) { struct inpcb *inp = mem; INP_LOCK_DESTROY(inp); } /* * in_pcbdrop() removes an inpcb from hashed lists, releasing its address and * port reservation, and preventing it from being returned by inpcb lookups. * * It is used by TCP to mark an inpcb as unused and avoid future packet * delivery or event notification when a socket remains open but TCP has * closed. This might occur as a result of a shutdown()-initiated TCP close * or a RST on the wire, and allows the port binding to be reused while still * maintaining the invariant that so_pcb always points to a valid inpcb until * in_pcbdetach(). * * XXXRW: Possibly in_pcbdrop() should also prevent future notifications by * in_pcbnotifyall() and in_pcbpurgeif0()? */ void in_pcbdrop(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); #ifdef INVARIANTS if (inp->inp_socket != NULL && inp->inp_ppcb != NULL) MPASS(inp->inp_refcount > 1); #endif inp->inp_flags |= INP_DROPPED; if (inp->inp_flags & INP_INHASHLIST) in_pcbremhash(inp); } #ifdef INET /* * Common routines to return the socket addresses associated with inpcbs. */ struct sockaddr * in_sockaddr(in_port_t port, struct in_addr *addr_p) { struct sockaddr_in *sin; sin = malloc(sizeof *sin, M_SONAME, M_WAITOK | M_ZERO); sin->sin_family = AF_INET; sin->sin_len = sizeof(*sin); sin->sin_addr = *addr_p; sin->sin_port = port; return (struct sockaddr *)sin; } int in_getsockaddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_getsockaddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_lport; addr = inp->inp_laddr; INP_RUNLOCK(inp); *nam = in_sockaddr(port, &addr); return 0; } int in_getpeeraddr(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; struct in_addr addr; in_port_t port; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_getpeeraddr: inp == NULL")); INP_RLOCK(inp); port = inp->inp_fport; addr = inp->inp_faddr; INP_RUNLOCK(inp); *nam = in_sockaddr(port, &addr); return 0; } void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr faddr, int errno, struct inpcb *(*notify)(struct inpcb *, int)) { struct inpcb *inp, *inp_temp; INP_INFO_WLOCK(pcbinfo); CK_LIST_FOREACH_SAFE(inp, &pcbinfo->ipi_listhead, inp_list, inp_temp) { INP_WLOCK(inp); #ifdef INET6 if ((inp->inp_vflag & INP_IPV4) == 0) { INP_WUNLOCK(inp); continue; } #endif if (inp->inp_faddr.s_addr != faddr.s_addr || inp->inp_socket == NULL) { INP_WUNLOCK(inp); continue; } if ((*notify)(inp, errno)) INP_WUNLOCK(inp); } INP_INFO_WUNLOCK(pcbinfo); } static bool inp_v4_multi_match(const struct inpcb *inp, void *v __unused) { if ((inp->inp_vflag & INP_IPV4) && inp->inp_moptions != NULL) return (true); else return (false); } void in_pcbpurgeif0(struct inpcbinfo *pcbinfo, struct ifnet *ifp) { struct inpcb_iterator inpi = INP_ITERATOR(pcbinfo, INPLOOKUP_WLOCKPCB, inp_v4_multi_match, NULL); struct inpcb *inp; struct in_multi *inm; struct in_mfilter *imf; struct ip_moptions *imo; IN_MULTI_LOCK_ASSERT(); while ((inp = inp_next(&inpi)) != NULL) { INP_WLOCK_ASSERT(inp); imo = inp->inp_moptions; /* * Unselect the outgoing interface if it is being * detached. */ if (imo->imo_multicast_ifp == ifp) imo->imo_multicast_ifp = NULL; /* * Drop multicast group membership if we joined * through the interface being detached. * * XXX This can all be deferred to an epoch_call */ restart: IP_MFILTER_FOREACH(imf, &imo->imo_head) { if ((inm = imf->imf_inm) == NULL) continue; if (inm->inm_ifp != ifp) continue; ip_mfilter_remove(&imo->imo_head, imf); in_leavegroup_locked(inm, NULL); ip_mfilter_free(imf); goto restart; } } } /* * Lookup a PCB based on the local address and port. Caller must hold the * hash lock. No inpcb locks or references are acquired. */ #define INP_LOOKUP_MAPPED_PCB_COST 3 struct inpcb * in_pcblookup_local(struct inpcbinfo *pcbinfo, struct in_addr laddr, u_short lport, int fib, int lookupflags, struct ucred *cred) { struct inpcb *inp; #ifdef INET6 int matchwild = 3 + INP_LOOKUP_MAPPED_PCB_COST; #else int matchwild = 3; #endif int wildcard; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); KASSERT(fib == RT_ALL_FIBS || (fib >= 0 && fib < V_rt_numfibs), ("%s: invalid fib %d", __func__, fib)); INP_HASH_LOCK_ASSERT(pcbinfo); if ((lookupflags & INPLOOKUP_WILDCARD) == 0) { struct inpcbhead *head; /* * Look for an unconnected (wildcard foreign addr) PCB that * matches the local address and port we're looking for. */ head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport, pcbinfo->ipi_hashmask)]; CK_LIST_FOREACH(inp, head, inp_hash_wild) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; #endif if (inp->inp_faddr.s_addr == INADDR_ANY && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_lport == lport && (fib == RT_ALL_FIBS || inp->inp_inc.inc_fibnum == fib)) { /* * Found? */ if (prison_equal_ip4(cred->cr_prison, inp->inp_cred->cr_prison)) return (inp); } } /* * Not found. */ return (NULL); } else { struct inpcbporthead *porthash; struct inpcbport *phd; struct inpcb *match = NULL; /* * Best fit PCB lookup. * * First see if this local port is in use by looking on the * port hash list. */ porthash = &pcbinfo->ipi_porthashbase[INP_PCBPORTHASH(lport, pcbinfo->ipi_porthashmask)]; CK_LIST_FOREACH(phd, porthash, phd_hash) { if (phd->phd_port == lport) break; } if (phd != NULL) { /* * Port is in use by one or more PCBs. Look for best * fit. */ CK_LIST_FOREACH(inp, &phd->phd_pcblist, inp_portlist) { wildcard = 0; if (!prison_equal_ip4(inp->inp_cred->cr_prison, cred->cr_prison)) continue; if (fib != RT_ALL_FIBS && inp->inp_inc.inc_fibnum != fib) continue; #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) continue; /* * We never select the PCB that has * INP_IPV6 flag and is bound to :: if * we have another PCB which is bound * to 0.0.0.0. If a PCB has the * INP_IPV6 flag, then we set its cost * higher than IPv4 only PCBs. * * Note that the case only happens * when a socket is bound to ::, under * the condition that the use of the * mapped address is allowed. */ if ((inp->inp_vflag & INP_IPV6) != 0) wildcard += INP_LOOKUP_MAPPED_PCB_COST; #endif if (inp->inp_faddr.s_addr != INADDR_ANY) wildcard++; if (inp->inp_laddr.s_addr != INADDR_ANY) { if (laddr.s_addr == INADDR_ANY) wildcard++; else if (inp->inp_laddr.s_addr != laddr.s_addr) continue; } else { if (laddr.s_addr != INADDR_ANY) wildcard++; } if (wildcard < matchwild) { match = inp; matchwild = wildcard; if (matchwild == 0) break; } } } return (match); } } #undef INP_LOOKUP_MAPPED_PCB_COST static bool in_pcblookup_lb_match(const struct inpcblbgroup *grp, int domain, int fib) { return ((domain == M_NODOM || domain == grp->il_numa_domain) && (fib == RT_ALL_FIBS || fib == grp->il_fibnum)); } static struct inpcb * in_pcblookup_lbgroup(const struct inpcbinfo *pcbinfo, const struct in_addr *faddr, uint16_t fport, const struct in_addr *laddr, uint16_t lport, int domain, int fib) { const struct inpcblbgrouphead *hdr; struct inpcblbgroup *grp; struct inpcblbgroup *jail_exact, *jail_wild, *local_exact, *local_wild; struct inpcb *inp; u_int count; INP_HASH_LOCK_ASSERT(pcbinfo); NET_EPOCH_ASSERT(); hdr = &pcbinfo->ipi_lbgrouphashbase[ INP_PCBPORTHASH(lport, pcbinfo->ipi_lbgrouphashmask)]; /* * Search for an LB group match based on the following criteria: * - prefer jailed groups to non-jailed groups * - prefer exact source address matches to wildcard matches * - prefer groups bound to the specified NUMA domain */ jail_exact = jail_wild = local_exact = local_wild = NULL; CK_LIST_FOREACH(grp, hdr, il_list) { bool injail; #ifdef INET6 if (!(grp->il_vflag & INP_IPV4)) continue; #endif if (grp->il_lport != lport) continue; injail = prison_flag(grp->il_cred, PR_IP4) != 0; if (injail && prison_check_ip4_locked(grp->il_cred->cr_prison, laddr) != 0) continue; if (grp->il_laddr.s_addr == laddr->s_addr) { if (injail) { jail_exact = grp; if (in_pcblookup_lb_match(grp, domain, fib)) /* This is a perfect match. */ goto out; } else if (local_exact == NULL || in_pcblookup_lb_match(grp, domain, fib)) { local_exact = grp; } } else if (grp->il_laddr.s_addr == INADDR_ANY) { if (injail) { if (jail_wild == NULL || in_pcblookup_lb_match(grp, domain, fib)) jail_wild = grp; } else if (local_wild == NULL || in_pcblookup_lb_match(grp, domain, fib)) { local_wild = grp; } } } if (jail_exact != NULL) grp = jail_exact; else if (jail_wild != NULL) grp = jail_wild; else if (local_exact != NULL) grp = local_exact; else grp = local_wild; if (grp == NULL) return (NULL); out: /* * Synchronize with in_pcblbgroup_insert(). */ count = atomic_load_acq_int(&grp->il_inpcnt); if (count == 0) return (NULL); inp = grp->il_inp[INP_PCBLBGROUP_PKTHASH(faddr, lport, fport) % count]; KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); return (inp); } static bool in_pcblookup_exact_match(const struct inpcb *inp, struct in_addr faddr, u_short fport, struct in_addr laddr, u_short lport) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) return (false); #endif if (inp->inp_faddr.s_addr == faddr.s_addr && inp->inp_laddr.s_addr == laddr.s_addr && inp->inp_fport == fport && inp->inp_lport == lport) return (true); return (false); } static struct inpcb * in_pcblookup_hash_exact(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_short fport, struct in_addr laddr, u_short lport) { struct inpcbhead *head; struct inpcb *inp; INP_HASH_LOCK_ASSERT(pcbinfo); head = &pcbinfo->ipi_hash_exact[INP_PCBHASH(&faddr, lport, fport, pcbinfo->ipi_hashmask)]; CK_LIST_FOREACH(inp, head, inp_hash_exact) { if (in_pcblookup_exact_match(inp, faddr, fport, laddr, lport)) return (inp); } return (NULL); } typedef enum { INPLOOKUP_MATCH_NONE = 0, INPLOOKUP_MATCH_WILD = 1, INPLOOKUP_MATCH_LADDR = 2, } inp_lookup_match_t; static inp_lookup_match_t in_pcblookup_wild_match(const struct inpcb *inp, struct in_addr laddr, u_short lport, int fib) { #ifdef INET6 /* XXX inp locking */ if ((inp->inp_vflag & INP_IPV4) == 0) return (INPLOOKUP_MATCH_NONE); #endif if (inp->inp_faddr.s_addr != INADDR_ANY || inp->inp_lport != lport) return (INPLOOKUP_MATCH_NONE); if (fib != RT_ALL_FIBS && inp->inp_inc.inc_fibnum != fib) return (INPLOOKUP_MATCH_NONE); if (inp->inp_laddr.s_addr == INADDR_ANY) return (INPLOOKUP_MATCH_WILD); if (inp->inp_laddr.s_addr == laddr.s_addr) return (INPLOOKUP_MATCH_LADDR); return (INPLOOKUP_MATCH_NONE); } #define INP_LOOKUP_AGAIN ((struct inpcb *)(uintptr_t)-1) static struct inpcb * in_pcblookup_hash_wild_smr(struct inpcbinfo *pcbinfo, struct in_addr laddr, u_short lport, int fib, const inp_lookup_t lockflags) { struct inpcbhead *head; struct inpcb *inp; KASSERT(SMR_ENTERED(pcbinfo->ipi_smr), ("%s: not in SMR read section", __func__)); head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport, pcbinfo->ipi_hashmask)]; CK_LIST_FOREACH(inp, head, inp_hash_wild) { inp_lookup_match_t match; match = in_pcblookup_wild_match(inp, laddr, lport, fib); if (match == INPLOOKUP_MATCH_NONE) continue; if (__predict_true(inp_smr_lock(inp, lockflags))) { match = in_pcblookup_wild_match(inp, laddr, lport, fib); if (match != INPLOOKUP_MATCH_NONE && prison_check_ip4_locked(inp->inp_cred->cr_prison, &laddr) == 0) return (inp); inp_unlock(inp, lockflags); } /* * The matching socket disappeared out from under us. Fall back * to a serialized lookup. */ return (INP_LOOKUP_AGAIN); } return (NULL); } static struct inpcb * in_pcblookup_hash_wild_locked(struct inpcbinfo *pcbinfo, struct in_addr laddr, u_short lport, int fib) { struct inpcbhead *head; struct inpcb *inp, *local_wild, *local_exact, *jail_wild; #ifdef INET6 struct inpcb *local_wild_mapped; #endif INP_HASH_LOCK_ASSERT(pcbinfo); /* * Order of socket selection - we always prefer jails. * 1. jailed, non-wild. * 2. jailed, wild. * 3. non-jailed, non-wild. * 4. non-jailed, wild. */ head = &pcbinfo->ipi_hash_wild[INP_PCBHASH_WILD(lport, pcbinfo->ipi_hashmask)]; local_wild = local_exact = jail_wild = NULL; #ifdef INET6 local_wild_mapped = NULL; #endif CK_LIST_FOREACH(inp, head, inp_hash_wild) { inp_lookup_match_t match; bool injail; match = in_pcblookup_wild_match(inp, laddr, lport, fib); if (match == INPLOOKUP_MATCH_NONE) continue; injail = prison_flag(inp->inp_cred, PR_IP4) != 0; if (injail) { if (prison_check_ip4_locked(inp->inp_cred->cr_prison, &laddr) != 0) continue; } else { if (local_exact != NULL) continue; } if (match == INPLOOKUP_MATCH_LADDR) { if (injail) return (inp); local_exact = inp; } else { #ifdef INET6 /* XXX inp locking, NULL check */ if (inp->inp_vflag & INP_IPV6PROTO) local_wild_mapped = inp; else #endif if (injail) jail_wild = inp; else local_wild = inp; } } if (jail_wild != NULL) return (jail_wild); if (local_exact != NULL) return (local_exact); if (local_wild != NULL) return (local_wild); #ifdef INET6 if (local_wild_mapped != NULL) return (local_wild_mapped); #endif return (NULL); } /* * Lookup PCB in hash list, using pcbinfo tables. This variation assumes * that the caller has either locked the hash list, which usually happens * for bind(2) operations, or is in SMR section, which happens when sorting * out incoming packets. */ static struct inpcb * in_pcblookup_hash_locked(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, uint8_t numa_domain, int fib) { struct inpcb *inp; const u_short fport = fport_arg, lport = lport_arg; KASSERT((lookupflags & ~(INPLOOKUP_WILDCARD | INPLOOKUP_FIB)) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); KASSERT(faddr.s_addr != INADDR_ANY, ("%s: invalid foreign address", __func__)); KASSERT(laddr.s_addr != INADDR_ANY, ("%s: invalid local address", __func__)); INP_HASH_WLOCK_ASSERT(pcbinfo); inp = in_pcblookup_hash_exact(pcbinfo, faddr, fport, laddr, lport); if (inp != NULL) return (inp); if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { inp = in_pcblookup_lbgroup(pcbinfo, &faddr, fport, &laddr, lport, numa_domain, fib); if (inp == NULL) { inp = in_pcblookup_hash_wild_locked(pcbinfo, laddr, lport, fib); } } return (inp); } static struct inpcb * in_pcblookup_hash(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, struct in_addr laddr, u_int lport, int lookupflags, uint8_t numa_domain, int fib) { struct inpcb *inp; const inp_lookup_t lockflags = lookupflags & INPLOOKUP_LOCKMASK; KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, ("%s: LOCKPCB not set", __func__)); INP_HASH_WLOCK(pcbinfo); inp = in_pcblookup_hash_locked(pcbinfo, faddr, fport, laddr, lport, lookupflags & ~INPLOOKUP_LOCKMASK, numa_domain, fib); if (inp != NULL && !inp_trylock(inp, lockflags)) { in_pcbref(inp); INP_HASH_WUNLOCK(pcbinfo); inp_lock(inp, lockflags); if (in_pcbrele(inp, lockflags)) /* XXX-MJ or retry until we get a negative match? */ inp = NULL; } else { INP_HASH_WUNLOCK(pcbinfo); } return (inp); } static struct inpcb * in_pcblookup_hash_smr(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport_arg, struct in_addr laddr, u_int lport_arg, int lookupflags, uint8_t numa_domain, int fib) { struct inpcb *inp; const inp_lookup_t lockflags = lookupflags & INPLOOKUP_LOCKMASK; const u_short fport = fport_arg, lport = lport_arg; KASSERT((lookupflags & ~INPLOOKUP_MASK) == 0, ("%s: invalid lookup flags %d", __func__, lookupflags)); KASSERT((lookupflags & (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB)) != 0, ("%s: LOCKPCB not set", __func__)); smr_enter(pcbinfo->ipi_smr); inp = in_pcblookup_hash_exact(pcbinfo, faddr, fport, laddr, lport); if (inp != NULL) { if (__predict_true(inp_smr_lock(inp, lockflags))) { /* * Revalidate the 4-tuple, the socket could have been * disconnected. */ if (__predict_true(in_pcblookup_exact_match(inp, faddr, fport, laddr, lport))) return (inp); inp_unlock(inp, lockflags); } /* * We failed to lock the inpcb, or its connection state changed * out from under us. Fall back to a precise search. */ return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, lookupflags, numa_domain, fib)); } if ((lookupflags & INPLOOKUP_WILDCARD) != 0) { inp = in_pcblookup_lbgroup(pcbinfo, &faddr, fport, &laddr, lport, numa_domain, fib); if (inp != NULL) { if (__predict_true(inp_smr_lock(inp, lockflags))) { if (__predict_true(in_pcblookup_wild_match(inp, laddr, lport, fib) != INPLOOKUP_MATCH_NONE)) return (inp); inp_unlock(inp, lockflags); } inp = INP_LOOKUP_AGAIN; } else { inp = in_pcblookup_hash_wild_smr(pcbinfo, laddr, lport, fib, lockflags); } if (inp == INP_LOOKUP_AGAIN) { return (in_pcblookup_hash(pcbinfo, faddr, fport, laddr, lport, lookupflags, numa_domain, fib)); } } if (inp == NULL) smr_exit(pcbinfo->ipi_smr); return (inp); } /* * Public inpcb lookup routines, accepting a 4-tuple, and optionally, an mbuf * from which a pre-calculated hash value may be extracted. */ struct inpcb * in_pcblookup(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp) { int fib; fib = (lookupflags & INPLOOKUP_FIB) ? if_getfib(ifp) : RT_ALL_FIBS; return (in_pcblookup_hash_smr(pcbinfo, faddr, fport, laddr, lport, lookupflags, M_NODOM, fib)); } struct inpcb * in_pcblookup_mbuf(struct inpcbinfo *pcbinfo, struct in_addr faddr, u_int fport, struct in_addr laddr, u_int lport, int lookupflags, struct ifnet *ifp __unused, struct mbuf *m) { int fib; M_ASSERTPKTHDR(m); fib = (lookupflags & INPLOOKUP_FIB) ? M_GETFIB(m) : RT_ALL_FIBS; return (in_pcblookup_hash_smr(pcbinfo, faddr, fport, laddr, lport, lookupflags, m->m_pkthdr.numa_domain, fib)); } #endif /* INET */ static bool in_pcbjailed(const struct inpcb *inp, unsigned int flag) { return (prison_flag(inp->inp_cred, flag) != 0); } /* * Insert the PCB into a hash chain using ordering rules which ensure that * in_pcblookup_hash_wild_*() always encounter the highest-ranking PCB first. * * Specifically, keep jailed PCBs in front of non-jailed PCBs, and keep PCBs * with exact local addresses ahead of wildcard PCBs. Unbound v4-mapped v6 PCBs * always appear last no matter whether they are jailed. */ static void _in_pcbinshash_wild(struct inpcbhead *pcbhash, struct inpcb *inp) { struct inpcb *last; bool bound, injail; INP_LOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); last = NULL; bound = inp->inp_laddr.s_addr != INADDR_ANY; if (!bound && (inp->inp_vflag & INP_IPV6PROTO) != 0) { CK_LIST_FOREACH(last, pcbhash, inp_hash_wild) { if (CK_LIST_NEXT(last, inp_hash_wild) == NULL) { CK_LIST_INSERT_AFTER(last, inp, inp_hash_wild); return; } } CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash_wild); return; } injail = in_pcbjailed(inp, PR_IP4); if (!injail) { CK_LIST_FOREACH(last, pcbhash, inp_hash_wild) { if (!in_pcbjailed(last, PR_IP4)) break; if (CK_LIST_NEXT(last, inp_hash_wild) == NULL) { CK_LIST_INSERT_AFTER(last, inp, inp_hash_wild); return; } } } else if (!CK_LIST_EMPTY(pcbhash) && !in_pcbjailed(CK_LIST_FIRST(pcbhash), PR_IP4)) { CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash_wild); return; } if (!bound) { CK_LIST_FOREACH_FROM(last, pcbhash, inp_hash_wild) { if (last->inp_laddr.s_addr == INADDR_ANY) break; if (CK_LIST_NEXT(last, inp_hash_wild) == NULL) { CK_LIST_INSERT_AFTER(last, inp, inp_hash_wild); return; } } } if (last == NULL) CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash_wild); else CK_LIST_INSERT_BEFORE(last, inp, inp_hash_wild); } #ifdef INET6 /* * See the comment above _in_pcbinshash_wild(). */ static void _in6_pcbinshash_wild(struct inpcbhead *pcbhash, struct inpcb *inp) { struct inpcb *last; bool bound, injail; INP_LOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); last = NULL; bound = !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr); injail = in_pcbjailed(inp, PR_IP6); if (!injail) { CK_LIST_FOREACH(last, pcbhash, inp_hash_wild) { if (!in_pcbjailed(last, PR_IP6)) break; if (CK_LIST_NEXT(last, inp_hash_wild) == NULL) { CK_LIST_INSERT_AFTER(last, inp, inp_hash_wild); return; } } } else if (!CK_LIST_EMPTY(pcbhash) && !in_pcbjailed(CK_LIST_FIRST(pcbhash), PR_IP6)) { CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash_wild); return; } if (!bound) { CK_LIST_FOREACH_FROM(last, pcbhash, inp_hash_wild) { if (IN6_IS_ADDR_UNSPECIFIED(&last->in6p_laddr)) break; if (CK_LIST_NEXT(last, inp_hash_wild) == NULL) { CK_LIST_INSERT_AFTER(last, inp, inp_hash_wild); return; } } } if (last == NULL) CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash_wild); else CK_LIST_INSERT_BEFORE(last, inp, inp_hash_wild); } #endif /* * Insert PCB onto various hash lists. */ int in_pcbinshash(struct inpcb *inp) { struct inpcbhead *pcbhash; struct inpcbporthead *pcbporthash; struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbport *phd; uint32_t hash; bool connected; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); KASSERT((inp->inp_flags & INP_INHASHLIST) == 0, ("in_pcbinshash: INP_INHASHLIST")); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { hash = INP6_PCBHASH(&inp->in6p_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask); connected = !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr); } else #endif { hash = INP_PCBHASH(&inp->inp_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask); connected = !in_nullhost(inp->inp_faddr); } if (connected) pcbhash = &pcbinfo->ipi_hash_exact[hash]; else pcbhash = &pcbinfo->ipi_hash_wild[hash]; pcbporthash = &pcbinfo->ipi_porthashbase[ INP_PCBPORTHASH(inp->inp_lport, pcbinfo->ipi_porthashmask)]; /* * Ignore SO_REUSEPORT_LB if the socket is connected. Really this case * should be an error, but for UDP sockets it is not, and some * applications erroneously set it on connected UDP sockets, so we can't * change this without breaking compatibility. */ if (!connected && (inp->inp_socket->so_options & SO_REUSEPORT_LB) != 0) { int error = in_pcbinslbgrouphash(inp, M_NODOM); if (error != 0) return (error); } /* * Go through port list and look for a head for this lport. */ CK_LIST_FOREACH(phd, pcbporthash, phd_hash) { if (phd->phd_port == inp->inp_lport) break; } /* * If none exists, malloc one and tack it on. */ if (phd == NULL) { phd = uma_zalloc_smr(pcbinfo->ipi_portzone, M_NOWAIT); if (phd == NULL) { if ((inp->inp_flags & INP_INLBGROUP) != 0) in_pcbremlbgrouphash(inp); return (ENOMEM); } phd->phd_port = inp->inp_lport; CK_LIST_INIT(&phd->phd_pcblist); CK_LIST_INSERT_HEAD(pcbporthash, phd, phd_hash); } inp->inp_phd = phd; CK_LIST_INSERT_HEAD(&phd->phd_pcblist, inp, inp_portlist); /* * The PCB may have been disconnected in the past. Before we can safely * make it visible in the hash table, we must wait for all readers which * may be traversing this PCB to finish. */ if (inp->inp_smr != SMR_SEQ_INVALID) { smr_wait(pcbinfo->ipi_smr, inp->inp_smr); inp->inp_smr = SMR_SEQ_INVALID; } if (connected) CK_LIST_INSERT_HEAD(pcbhash, inp, inp_hash_exact); else { #ifdef INET6 if ((inp->inp_vflag & INP_IPV6) != 0) _in6_pcbinshash_wild(pcbhash, inp); else #endif _in_pcbinshash_wild(pcbhash, inp); } inp->inp_flags |= INP_INHASHLIST; return (0); } void in_pcbremhash_locked(struct inpcb *inp) { struct inpcbport *phd = inp->inp_phd; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(inp->inp_pcbinfo); MPASS(inp->inp_flags & INP_INHASHLIST); if ((inp->inp_flags & INP_INLBGROUP) != 0) in_pcbremlbgrouphash(inp); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr)) CK_LIST_REMOVE(inp, inp_hash_wild); else CK_LIST_REMOVE(inp, inp_hash_exact); } else #endif { if (in_nullhost(inp->inp_faddr)) CK_LIST_REMOVE(inp, inp_hash_wild); else CK_LIST_REMOVE(inp, inp_hash_exact); } CK_LIST_REMOVE(inp, inp_portlist); if (CK_LIST_FIRST(&phd->phd_pcblist) == NULL) { CK_LIST_REMOVE(phd, phd_hash); uma_zfree_smr(inp->inp_pcbinfo->ipi_portzone, phd); } inp->inp_flags &= ~INP_INHASHLIST; } static void in_pcbremhash(struct inpcb *inp) { INP_HASH_WLOCK(inp->inp_pcbinfo); in_pcbremhash_locked(inp); INP_HASH_WUNLOCK(inp->inp_pcbinfo); } /* * Move PCB to the proper hash bucket when { faddr, fport } have been * changed. NOTE: This does not handle the case of the lport changing (the * hashed port list would have to be updated as well), so the lport must * not change after in_pcbinshash() has been called. */ void in_pcbrehash(struct inpcb *inp) { struct inpcbinfo *pcbinfo = inp->inp_pcbinfo; struct inpcbhead *head; uint32_t hash; bool connected; INP_WLOCK_ASSERT(inp); INP_HASH_WLOCK_ASSERT(pcbinfo); KASSERT(inp->inp_flags & INP_INHASHLIST, ("%s: !INP_INHASHLIST", __func__)); KASSERT(inp->inp_smr == SMR_SEQ_INVALID, ("%s: inp was disconnected", __func__)); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { hash = INP6_PCBHASH(&inp->in6p_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask); connected = !IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_faddr); } else #endif { hash = INP_PCBHASH(&inp->inp_faddr, inp->inp_lport, inp->inp_fport, pcbinfo->ipi_hashmask); connected = !in_nullhost(inp->inp_faddr); } /* See the comment in in_pcbinshash(). */ if (connected && (inp->inp_flags & INP_INLBGROUP) != 0) in_pcbremlbgrouphash(inp); /* * When rehashing, the caller must ensure that either the new or the old * foreign address was unspecified. */ if (connected) CK_LIST_REMOVE(inp, inp_hash_wild); else CK_LIST_REMOVE(inp, inp_hash_exact); if (connected) { head = &pcbinfo->ipi_hash_exact[hash]; CK_LIST_INSERT_HEAD(head, inp, inp_hash_exact); } else { head = &pcbinfo->ipi_hash_wild[hash]; CK_LIST_INSERT_HEAD(head, inp, inp_hash_wild); } } /* * Check for alternatives when higher level complains * about service problems. For now, invalidate cached * routing information. If the route was created dynamically * (by a redirect), time to try a default gateway again. */ void in_losing(struct inpcb *inp) { RO_INVALIDATE_CACHE(&inp->inp_route); return; } /* * A set label operation has occurred at the socket layer, propagate the * label change into the in_pcb for the socket. */ void in_pcbsosetlabel(struct socket *so) { #ifdef MAC struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("in_pcbsosetlabel: so->so_pcb == NULL")); INP_WLOCK(inp); SOCK_LOCK(so); mac_inpcb_sosetlabel(so, inp); SOCK_UNLOCK(so); INP_WUNLOCK(inp); #endif } void inp_wlock(struct inpcb *inp) { INP_WLOCK(inp); } void inp_wunlock(struct inpcb *inp) { INP_WUNLOCK(inp); } void inp_rlock(struct inpcb *inp) { INP_RLOCK(inp); } void inp_runlock(struct inpcb *inp) { INP_RUNLOCK(inp); } #ifdef INVARIANT_SUPPORT void inp_lock_assert(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); } void inp_unlock_assert(struct inpcb *inp) { INP_UNLOCK_ASSERT(inp); } #endif void inp_apply_all(struct inpcbinfo *pcbinfo, void (*func)(struct inpcb *, void *), void *arg) { struct inpcb_iterator inpi = INP_ALL_ITERATOR(pcbinfo, INPLOOKUP_WLOCKPCB); struct inpcb *inp; while ((inp = inp_next(&inpi)) != NULL) func(inp, arg); } struct socket * inp_inpcbtosocket(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); return (inp->inp_socket); } struct tcpcb * inp_inpcbtotcpcb(struct inpcb *inp) { INP_WLOCK_ASSERT(inp); return ((struct tcpcb *)inp->inp_ppcb); } int inp_ip_tos_get(const struct inpcb *inp) { return (inp->inp_ip_tos); } void inp_ip_tos_set(struct inpcb *inp, int val) { inp->inp_ip_tos = val; } void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, uint32_t *faddr, uint16_t *fp) { INP_LOCK_ASSERT(inp); *laddr = inp->inp_laddr.s_addr; *faddr = inp->inp_faddr.s_addr; *lp = inp->inp_lport; *fp = inp->inp_fport; } struct inpcb * so_sotoinpcb(struct socket *so) { return (sotoinpcb(so)); } /* * Create an external-format (``xinpcb'') structure using the information in * the kernel-format in_pcb structure pointed to by inp. This is done to * reduce the spew of irrelevant information over this interface, to isolate * user code from changes in the kernel structure, and potentially to provide * information-hiding if we decide that some of this information should be * hidden from users. */ void in_pcbtoxinpcb(const struct inpcb *inp, struct xinpcb *xi) { bzero(xi, sizeof(*xi)); xi->xi_len = sizeof(struct xinpcb); if (inp->inp_socket) sotoxsocket(inp->inp_socket, &xi->xi_socket); bcopy(&inp->inp_inc, &xi->inp_inc, sizeof(struct in_conninfo)); xi->inp_gencnt = inp->inp_gencnt; xi->inp_ppcb = (uintptr_t)inp->inp_ppcb; xi->inp_flow = inp->inp_flow; xi->inp_flowid = inp->inp_flowid; xi->inp_flowtype = inp->inp_flowtype; xi->inp_flags = inp->inp_flags; xi->inp_flags2 = inp->inp_flags2; xi->in6p_cksum = inp->in6p_cksum; xi->in6p_hops = inp->in6p_hops; xi->inp_ip_tos = inp->inp_ip_tos; xi->inp_vflag = inp->inp_vflag; xi->inp_ip_ttl = inp->inp_ip_ttl; xi->inp_ip_p = inp->inp_ip_p; xi->inp_ip_minttl = inp->inp_ip_minttl; } int sysctl_setsockopt(SYSCTL_HANDLER_ARGS, struct inpcbinfo *pcbinfo, int (*ctloutput_set)(struct inpcb *, struct sockopt *)) { struct sockopt sopt; struct inpcb_iterator inpi = INP_ALL_ITERATOR(pcbinfo, INPLOOKUP_WLOCKPCB); struct inpcb *inp; struct sockopt_parameters *params; struct socket *so; int error; char buf[1024]; if (req->oldptr != NULL || req->oldlen != 0) return (EINVAL); if (req->newptr == NULL) return (EPERM); if (req->newlen > sizeof(buf)) return (ENOMEM); error = SYSCTL_IN(req, buf, req->newlen); if (error != 0) return (error); if (req->newlen < sizeof(struct sockopt_parameters)) return (EINVAL); params = (struct sockopt_parameters *)buf; sopt.sopt_level = params->sop_level; sopt.sopt_name = params->sop_optname; sopt.sopt_dir = SOPT_SET; sopt.sopt_val = params->sop_optval; sopt.sopt_valsize = req->newlen - sizeof(struct sockopt_parameters); sopt.sopt_td = NULL; #ifdef INET6 if (params->sop_inc.inc_flags & INC_ISIPV6) { if (IN6_IS_SCOPE_LINKLOCAL(¶ms->sop_inc.inc6_laddr)) params->sop_inc.inc6_laddr.s6_addr16[1] = htons(params->sop_inc.inc6_zoneid & 0xffff); if (IN6_IS_SCOPE_LINKLOCAL(¶ms->sop_inc.inc6_faddr)) params->sop_inc.inc6_faddr.s6_addr16[1] = htons(params->sop_inc.inc6_zoneid & 0xffff); } #endif if (params->sop_inc.inc_lport != htons(0) && params->sop_inc.inc_fport != htons(0)) { #ifdef INET6 if (params->sop_inc.inc_flags & INC_ISIPV6) inpi.hash = INP6_PCBHASH( ¶ms->sop_inc.inc6_faddr, params->sop_inc.inc_lport, params->sop_inc.inc_fport, pcbinfo->ipi_hashmask); else #endif inpi.hash = INP_PCBHASH( ¶ms->sop_inc.inc_faddr, params->sop_inc.inc_lport, params->sop_inc.inc_fport, pcbinfo->ipi_hashmask); } while ((inp = inp_next(&inpi)) != NULL) if (inp->inp_gencnt == params->sop_id) { if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (ECONNRESET); } so = inp->inp_socket; KASSERT(so != NULL, ("inp_socket == NULL")); soref(so); if (params->sop_level == SOL_SOCKET) { INP_WUNLOCK(inp); error = sosetopt(so, &sopt); } else error = (*ctloutput_set)(inp, &sopt); sorele(so); break; } if (inp == NULL) error = ESRCH; return (error); } #ifdef DDB static void db_print_indent(int indent) { int i; for (i = 0; i < indent; i++) db_printf(" "); } static void db_print_inconninfo(struct in_conninfo *inc, const char *name, int indent) { char faddr_str[48], laddr_str[48]; db_print_indent(indent); db_printf("%s at %p\n", name, inc); indent += 2; #ifdef INET6 if (inc->inc_flags & INC_ISIPV6) { /* IPv6. */ ip6_sprintf(laddr_str, &inc->inc6_laddr); ip6_sprintf(faddr_str, &inc->inc6_faddr); } else #endif { /* IPv4. */ inet_ntoa_r(inc->inc_laddr, laddr_str); inet_ntoa_r(inc->inc_faddr, faddr_str); } db_print_indent(indent); db_printf("inc_laddr %s inc_lport %u\n", laddr_str, ntohs(inc->inc_lport)); db_print_indent(indent); db_printf("inc_faddr %s inc_fport %u\n", faddr_str, ntohs(inc->inc_fport)); } -static void +void db_print_inpflags(int inp_flags) { int comma; comma = 0; if (inp_flags & INP_RECVOPTS) { db_printf("%sINP_RECVOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVRETOPTS) { db_printf("%sINP_RECVRETOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVDSTADDR) { db_printf("%sINP_RECVDSTADDR", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ORIGDSTADDR) { db_printf("%sINP_ORIGDSTADDR", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_HDRINCL) { db_printf("%sINP_HDRINCL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_HIGHPORT) { db_printf("%sINP_HIGHPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_LOWPORT) { db_printf("%sINP_LOWPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ANONPORT) { db_printf("%sINP_ANONPORT", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVIF) { db_printf("%sINP_RECVIF", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_MTUDISC) { db_printf("%sINP_MTUDISC", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVTTL) { db_printf("%sINP_RECVTTL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_DONTFRAG) { db_printf("%sINP_DONTFRAG", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_RECVTOS) { db_printf("%sINP_RECVTOS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_IPV6_V6ONLY) { db_printf("%sIN6P_IPV6_V6ONLY", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_PKTINFO) { db_printf("%sIN6P_PKTINFO", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_HOPLIMIT) { db_printf("%sIN6P_HOPLIMIT", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_HOPOPTS) { db_printf("%sIN6P_HOPOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_DSTOPTS) { db_printf("%sIN6P_DSTOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RTHDR) { db_printf("%sIN6P_RTHDR", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RTHDRDSTOPTS) { db_printf("%sIN6P_RTHDRDSTOPTS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_TCLASS) { db_printf("%sIN6P_TCLASS", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_AUTOFLOWLABEL) { db_printf("%sIN6P_AUTOFLOWLABEL", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_ONESBCAST) { db_printf("%sINP_ONESBCAST", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_DROPPED) { db_printf("%sINP_DROPPED", comma ? ", " : ""); comma = 1; } if (inp_flags & INP_SOCKREF) { db_printf("%sINP_SOCKREF", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_RFC2292) { db_printf("%sIN6P_RFC2292", comma ? ", " : ""); comma = 1; } if (inp_flags & IN6P_MTU) { db_printf("IN6P_MTU%s", comma ? ", " : ""); comma = 1; } } static void db_print_inpvflag(u_char inp_vflag) { int comma; comma = 0; if (inp_vflag & INP_IPV4) { db_printf("%sINP_IPV4", comma ? ", " : ""); comma = 1; } if (inp_vflag & INP_IPV6) { db_printf("%sINP_IPV6", comma ? ", " : ""); comma = 1; } if (inp_vflag & INP_IPV6PROTO) { db_printf("%sINP_IPV6PROTO", comma ? ", " : ""); comma = 1; } } static void db_print_inpcb(struct inpcb *inp, const char *name, int indent) { db_print_indent(indent); db_printf("%s at %p\n", name, inp); indent += 2; db_print_indent(indent); db_printf("inp_flow: 0x%x\n", inp->inp_flow); db_print_inconninfo(&inp->inp_inc, "inp_conninfo", indent); db_print_indent(indent); db_printf("inp_ppcb: %p inp_pcbinfo: %p inp_socket: %p\n", inp->inp_ppcb, inp->inp_pcbinfo, inp->inp_socket); db_print_indent(indent); db_printf("inp_label: %p inp_flags: 0x%x (", inp->inp_label, inp->inp_flags); db_print_inpflags(inp->inp_flags); db_printf(")\n"); db_print_indent(indent); db_printf("inp_sp: %p inp_vflag: 0x%x (", inp->inp_sp, inp->inp_vflag); db_print_inpvflag(inp->inp_vflag); db_printf(")\n"); db_print_indent(indent); db_printf("inp_ip_ttl: %d inp_ip_p: %d inp_ip_minttl: %d\n", inp->inp_ip_ttl, inp->inp_ip_p, inp->inp_ip_minttl); db_print_indent(indent); #ifdef INET6 if (inp->inp_vflag & INP_IPV6) { db_printf("in6p_options: %p in6p_outputopts: %p " "in6p_moptions: %p\n", inp->in6p_options, inp->in6p_outputopts, inp->in6p_moptions); db_printf("in6p_icmp6filt: %p in6p_cksum %d " "in6p_hops %u\n", inp->in6p_icmp6filt, inp->in6p_cksum, inp->in6p_hops); } else #endif { db_printf("inp_ip_tos: %d inp_ip_options: %p " "inp_ip_moptions: %p\n", inp->inp_ip_tos, inp->inp_options, inp->inp_moptions); } db_print_indent(indent); db_printf("inp_phd: %p inp_gencnt: %ju\n", inp->inp_phd, (uintmax_t)inp->inp_gencnt); } DB_SHOW_COMMAND(inpcb, db_show_inpcb) { struct inpcb *inp; if (!have_addr) { db_printf("usage: show inpcb \n"); return; } inp = (struct inpcb *)addr; db_print_inpcb(inp, "inpcb", 0); } #endif /* DDB */ #ifdef RATELIMIT /* * Modify TX rate limit based on the existing "inp->inp_snd_tag", * if any. */ int in_pcbmodify_txrtlmt(struct inpcb *inp, uint32_t max_pacing_rate) { union if_snd_tag_modify_params params = { .rate_limit.max_rate = max_pacing_rate, .rate_limit.flags = M_NOWAIT, }; struct m_snd_tag *mst; int error; mst = inp->inp_snd_tag; if (mst == NULL) return (EINVAL); if (mst->sw->snd_tag_modify == NULL) { error = EOPNOTSUPP; } else { error = mst->sw->snd_tag_modify(mst, ¶ms); } return (error); } /* * Query existing TX rate limit based on the existing * "inp->inp_snd_tag", if any. */ int in_pcbquery_txrtlmt(struct inpcb *inp, uint32_t *p_max_pacing_rate) { union if_snd_tag_query_params params = { }; struct m_snd_tag *mst; int error; mst = inp->inp_snd_tag; if (mst == NULL) return (EINVAL); if (mst->sw->snd_tag_query == NULL) { error = EOPNOTSUPP; } else { error = mst->sw->snd_tag_query(mst, ¶ms); if (error == 0 && p_max_pacing_rate != NULL) *p_max_pacing_rate = params.rate_limit.max_rate; } return (error); } /* * Query existing TX queue level based on the existing * "inp->inp_snd_tag", if any. */ int in_pcbquery_txrlevel(struct inpcb *inp, uint32_t *p_txqueue_level) { union if_snd_tag_query_params params = { }; struct m_snd_tag *mst; int error; mst = inp->inp_snd_tag; if (mst == NULL) return (EINVAL); if (mst->sw->snd_tag_query == NULL) return (EOPNOTSUPP); error = mst->sw->snd_tag_query(mst, ¶ms); if (error == 0 && p_txqueue_level != NULL) *p_txqueue_level = params.rate_limit.queue_level; return (error); } /* * Allocate a new TX rate limit send tag from the network interface * given by the "ifp" argument and save it in "inp->inp_snd_tag": */ int in_pcbattach_txrtlmt(struct inpcb *inp, struct ifnet *ifp, uint32_t flowtype, uint32_t flowid, uint32_t max_pacing_rate, struct m_snd_tag **st) { union if_snd_tag_alloc_params params = { .rate_limit.hdr.type = (max_pacing_rate == -1U) ? IF_SND_TAG_TYPE_UNLIMITED : IF_SND_TAG_TYPE_RATE_LIMIT, .rate_limit.hdr.flowid = flowid, .rate_limit.hdr.flowtype = flowtype, .rate_limit.hdr.numa_domain = inp->inp_numa_domain, .rate_limit.max_rate = max_pacing_rate, .rate_limit.flags = M_NOWAIT, }; int error; INP_WLOCK_ASSERT(inp); /* * If there is already a send tag, or the INP is being torn * down, allocating a new send tag is not allowed. Else send * tags may leak. */ if (*st != NULL || (inp->inp_flags & INP_DROPPED) != 0) return (EINVAL); error = m_snd_tag_alloc(ifp, ¶ms, st); #ifdef INET if (error == 0) { counter_u64_add(rate_limit_set_ok, 1); counter_u64_add(rate_limit_active, 1); } else if (error != EOPNOTSUPP) counter_u64_add(rate_limit_alloc_fail, 1); #endif return (error); } void in_pcbdetach_tag(struct m_snd_tag *mst) { m_snd_tag_rele(mst); #ifdef INET counter_u64_add(rate_limit_active, -1); #endif } /* * Free an existing TX rate limit tag based on the "inp->inp_snd_tag", * if any: */ void in_pcbdetach_txrtlmt(struct inpcb *inp) { struct m_snd_tag *mst; INP_WLOCK_ASSERT(inp); mst = inp->inp_snd_tag; inp->inp_snd_tag = NULL; if (mst == NULL) return; m_snd_tag_rele(mst); #ifdef INET counter_u64_add(rate_limit_active, -1); #endif } int in_pcboutput_txrtlmt_locked(struct inpcb *inp, struct ifnet *ifp, struct mbuf *mb, uint32_t max_pacing_rate) { int error; /* * If the existing send tag is for the wrong interface due to * a route change, first drop the existing tag. Set the * CHANGED flag so that we will keep trying to allocate a new * tag if we fail to allocate one this time. */ if (inp->inp_snd_tag != NULL && inp->inp_snd_tag->ifp != ifp) { in_pcbdetach_txrtlmt(inp); inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED; } /* * NOTE: When attaching to a network interface a reference is * made to ensure the network interface doesn't go away until * all ratelimit connections are gone. The network interface * pointers compared below represent valid network interfaces, * except when comparing towards NULL. */ if (max_pacing_rate == 0 && inp->inp_snd_tag == NULL) { error = 0; } else if (!(ifp->if_capenable & IFCAP_TXRTLMT)) { if (inp->inp_snd_tag != NULL) in_pcbdetach_txrtlmt(inp); error = 0; } else if (inp->inp_snd_tag == NULL) { /* * In order to utilize packet pacing with RSS, we need * to wait until there is a valid RSS hash before we * can proceed: */ if (M_HASHTYPE_GET(mb) == M_HASHTYPE_NONE) { error = EAGAIN; } else { error = in_pcbattach_txrtlmt(inp, ifp, M_HASHTYPE_GET(mb), mb->m_pkthdr.flowid, max_pacing_rate, &inp->inp_snd_tag); } } else { error = in_pcbmodify_txrtlmt(inp, max_pacing_rate); } if (error == 0 || error == EOPNOTSUPP) inp->inp_flags2 &= ~INP_RATE_LIMIT_CHANGED; return (error); } /* * This function should be called when the INP_RATE_LIMIT_CHANGED flag * is set in the fast path and will attach/detach/modify the TX rate * limit send tag based on the socket's so_max_pacing_rate value. */ void in_pcboutput_txrtlmt(struct inpcb *inp, struct ifnet *ifp, struct mbuf *mb) { struct socket *socket; uint32_t max_pacing_rate; bool did_upgrade; if (inp == NULL) return; socket = inp->inp_socket; if (socket == NULL) return; if (!INP_WLOCKED(inp)) { /* * NOTE: If the write locking fails, we need to bail * out and use the non-ratelimited ring for the * transmit until there is a new chance to get the * write lock. */ if (!INP_TRY_UPGRADE(inp)) return; did_upgrade = 1; } else { did_upgrade = 0; } /* * NOTE: The so_max_pacing_rate value is read unlocked, * because atomic updates are not required since the variable * is checked at every mbuf we send. It is assumed that the * variable read itself will be atomic. */ max_pacing_rate = socket->so_max_pacing_rate; in_pcboutput_txrtlmt_locked(inp, ifp, mb, max_pacing_rate); if (did_upgrade) INP_DOWNGRADE(inp); } /* * Track route changes for TX rate limiting. */ void in_pcboutput_eagain(struct inpcb *inp) { bool did_upgrade; if (inp == NULL) return; if (inp->inp_snd_tag == NULL) return; if (!INP_WLOCKED(inp)) { /* * NOTE: If the write locking fails, we need to bail * out and use the non-ratelimited ring for the * transmit until there is a new chance to get the * write lock. */ if (!INP_TRY_UPGRADE(inp)) return; did_upgrade = 1; } else { did_upgrade = 0; } /* detach rate limiting */ in_pcbdetach_txrtlmt(inp); /* make sure new mbuf send tag allocation is made */ inp->inp_flags2 |= INP_RATE_LIMIT_CHANGED; if (did_upgrade) INP_DOWNGRADE(inp); } #ifdef INET static void rl_init(void *st) { rate_limit_new = counter_u64_alloc(M_WAITOK); rate_limit_chg = counter_u64_alloc(M_WAITOK); rate_limit_active = counter_u64_alloc(M_WAITOK); rate_limit_alloc_fail = counter_u64_alloc(M_WAITOK); rate_limit_set_ok = counter_u64_alloc(M_WAITOK); } SYSINIT(rl, SI_SUB_PROTO_DOMAININIT, SI_ORDER_ANY, rl_init, NULL); #endif #endif /* RATELIMIT */ diff --git a/sys/netinet/in_pcb.h b/sys/netinet/in_pcb.h index 17f24ba0bee7..9095e6bb71bf 100644 --- a/sys/netinet/in_pcb.h +++ b/sys/netinet/in_pcb.h @@ -1,728 +1,731 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1990, 1993 * The Regents of the University of California. * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * @(#)in_pcb.h 8.1 (Berkeley) 6/10/93 */ #ifndef _NETINET_IN_PCB_H_ #define _NETINET_IN_PCB_H_ #include #include #include #include #include #include #include #ifdef _KERNEL #include #include #include #include #include #include #endif #include /* * struct inpcb is the common protocol control block structure used in most * IP transport protocols. * * Pointers to local and foreign host table entries, local and foreign socket * numbers, and pointers up (to a socket structure) and down (to a * protocol-specific control block) are stored here. */ CK_LIST_HEAD(inpcbhead, inpcb); CK_LIST_HEAD(inpcbporthead, inpcbport); CK_LIST_HEAD(inpcblbgrouphead, inpcblbgroup); typedef uint64_t inp_gen_t; /* * PCB with AF_INET6 null bind'ed laddr can receive AF_INET input packet. * So, AF_INET6 null laddr is also used as AF_INET null laddr, by utilizing * the following structure. This requires padding always be zeroed out, * which is done right after inpcb allocation and stays through its lifetime. */ struct in_addr_4in6 { u_int32_t ia46_pad32[3]; struct in_addr ia46_addr4; }; union in_dependaddr { struct in_addr_4in6 id46_addr; struct in6_addr id6_addr; }; /* * NOTE: ipv6 addrs should be 64-bit aligned, per RFC 2553. in_conninfo has * some extra padding to accomplish this. * NOTE 2: tcp_syncache.c uses first 5 32-bit words, which identify fport, * lport, faddr to generate hash, so these fields shouldn't be moved. */ struct in_endpoints { u_int16_t ie_fport; /* foreign port */ u_int16_t ie_lport; /* local port */ /* protocol dependent part, local and foreign addr */ union in_dependaddr ie_dependfaddr; /* foreign host table entry */ union in_dependaddr ie_dependladdr; /* local host table entry */ #define ie_faddr ie_dependfaddr.id46_addr.ia46_addr4 #define ie_laddr ie_dependladdr.id46_addr.ia46_addr4 #define ie6_faddr ie_dependfaddr.id6_addr #define ie6_laddr ie_dependladdr.id6_addr u_int32_t ie6_zoneid; /* scope zone id */ }; /* * XXX The defines for inc_* are hacks and should be changed to direct * references. */ struct in_conninfo { u_int8_t inc_flags; u_int8_t inc_len; u_int16_t inc_fibnum; /* XXX was pad, 16 bits is plenty */ /* protocol dependent part */ struct in_endpoints inc_ie; }; /* * Flags for inc_flags. */ #define INC_ISIPV6 0x01 #define INC_IPV6MINMTU 0x02 #define inc_fport inc_ie.ie_fport #define inc_lport inc_ie.ie_lport #define inc_faddr inc_ie.ie_faddr #define inc_laddr inc_ie.ie_laddr #define inc6_faddr inc_ie.ie6_faddr #define inc6_laddr inc_ie.ie6_laddr #define inc6_zoneid inc_ie.ie6_zoneid #if defined(_KERNEL) || defined(_WANT_INPCB) /* * struct inpcb captures the network layer state for TCP, UDP, and raw IPv4 and * IPv6 sockets. In the case of TCP and UDP, further per-connection state is * hung off of inp_ppcb most of the time. Almost all fields of struct inpcb * are static after creation or protected by a per-inpcb rwlock, inp_lock. * * A inpcb database is indexed by addresses/ports hash as well as list of * all pcbs that belong to a certain proto. Database lookups or list traversals * are be performed inside SMR section. Once desired PCB is found its own * lock is to be obtained and SMR section exited. * * Key: * (c) - Constant after initialization * (e) - Protected by the SMR section * (i) - Protected by the inpcb lock * (p) - Protected by the pcbinfo lock for the inpcb * (h) - Protected by the pcbhash lock for the inpcb * (s) - Protected by another subsystem's locks * (x) - Undefined locking * * A few other notes: * * When a read lock is held, stability of the field is guaranteed; to write * to a field, a write lock must generally be held. * * netinet/netinet6-layer code should not assume that the inp_socket pointer * is safe to dereference without inp_lock being held, there may be * close(2)-related races. * * The inp_vflag field is overloaded, and would otherwise ideally be (c). */ struct icmp6_filter; struct inpcbpolicy; struct m_snd_tag; struct inpcb { /* Cache line #1 (amd64) */ CK_LIST_ENTRY(inpcb) inp_hash_exact; /* hash table linkage */ CK_LIST_ENTRY(inpcb) inp_hash_wild; /* hash table linkage */ struct rwlock inp_lock; /* Cache line #2 (amd64) */ #define inp_start_zero inp_refcount #define inp_zero_size (sizeof(struct inpcb) - \ offsetof(struct inpcb, inp_start_zero)) u_int inp_refcount; /* (i) refcount */ int inp_flags; /* (i) generic IP/datagram flags */ int inp_flags2; /* (i) generic IP/datagram flags #2*/ uint8_t inp_numa_domain; /* numa domain */ void *inp_ppcb; /* (i) pointer to per-protocol pcb */ struct socket *inp_socket; /* (i) back pointer to socket */ struct inpcbinfo *inp_pcbinfo; /* (c) PCB list info */ struct ucred *inp_cred; /* (c) cache of socket cred */ u_int32_t inp_flow; /* (i) IPv6 flow information */ u_char inp_vflag; /* (i) IP version flag (v4/v6) */ u_char inp_ip_ttl; /* (i) time to live proto */ u_char inp_ip_p; /* (c) protocol proto */ u_char inp_ip_minttl; /* (i) minimum TTL or drop */ uint32_t inp_flowid; /* (x) flow id / queue id */ smr_seq_t inp_smr; /* (i) sequence number at disconnect */ struct m_snd_tag *inp_snd_tag; /* (i) send tag for outgoing mbufs */ uint32_t inp_flowtype; /* (x) M_HASHTYPE value */ /* Local and foreign ports, local and foreign addr. */ struct in_conninfo inp_inc; /* (i,h) list for PCB's local port */ /* MAC and IPSEC policy information. */ struct label *inp_label; /* (i) MAC label */ struct inpcbpolicy *inp_sp; /* (s) for IPSEC */ /* Protocol-dependent part; options. */ struct { u_char inp_ip_tos; /* (i) type of service proto */ struct mbuf *inp_options; /* (i) IP options */ struct ip_moptions *inp_moptions; /* (i) mcast options */ }; struct { /* (i) IP options */ struct mbuf *in6p_options; /* (i) IP6 options for outgoing packets */ struct ip6_pktopts *in6p_outputopts; /* (i) IP multicast options */ struct ip6_moptions *in6p_moptions; /* (i) ICMPv6 code type filter */ struct icmp6_filter *in6p_icmp6filt; /* (i) IPV6_CHECKSUM setsockopt */ int in6p_cksum; short in6p_hops; }; CK_LIST_ENTRY(inpcb) inp_portlist; /* (r:e/w:h) port list */ struct inpcbport *inp_phd; /* (r:e/w:h) head of this list */ inp_gen_t inp_gencnt; /* (c) generation count */ void *spare_ptr; /* Spare pointer. */ rt_gen_t inp_rt_cookie; /* generation for route entry */ union { /* cached L3 information */ struct route inp_route; struct route_in6 inp_route6; }; CK_LIST_ENTRY(inpcb) inp_list; /* (r:e/w:p) all PCBs for proto */ }; #endif /* _KERNEL */ #define inp_fport inp_inc.inc_fport #define inp_lport inp_inc.inc_lport #define inp_faddr inp_inc.inc_faddr #define inp_laddr inp_inc.inc_laddr #define in6p_faddr inp_inc.inc6_faddr #define in6p_laddr inp_inc.inc6_laddr #define in6p_zoneid inp_inc.inc6_zoneid #define inp_vnet inp_pcbinfo->ipi_vnet /* * The range of the generation count, as used in this implementation, is 9e19. * We would have to create 300 billion connections per second for this number * to roll over in a year. This seems sufficiently unlikely that we simply * don't concern ourselves with that possibility. */ /* * Interface exported to userland by various protocols which use inpcbs. Hack * alert -- only define if struct xsocket is in scope. * Fields prefixed with "xi_" are unique to this structure, and the rest * match fields in the struct inpcb, to ease coding and porting. * * Legend: * (s) - used by userland utilities in src * (p) - used by utilities in ports * (3) - is known to be used by third party software not in ports * (n) - no known usage */ #ifdef _SYS_SOCKETVAR_H_ struct xinpcb { ksize_t xi_len; /* length of this structure */ struct xsocket xi_socket; /* (s,p) */ struct in_conninfo inp_inc; /* (s,p) */ uint64_t inp_gencnt; /* (s,p) */ kvaddr_t inp_ppcb; /* (s) netstat(1) */ int64_t inp_spare64[4]; uint32_t inp_flow; /* (s) */ uint32_t inp_flowid; /* (s) */ uint32_t inp_flowtype; /* (s) */ int32_t inp_flags; /* (s,p) */ int32_t inp_flags2; /* (s) */ uint32_t inp_unused; int32_t in6p_cksum; /* (n) */ int32_t inp_spare32[4]; uint16_t in6p_hops; /* (n) */ uint8_t inp_ip_tos; /* (n) */ int8_t pad8; uint8_t inp_vflag; /* (s,p) */ uint8_t inp_ip_ttl; /* (n) */ uint8_t inp_ip_p; /* (n) */ uint8_t inp_ip_minttl; /* (n) */ int8_t inp_spare8[4]; } __aligned(8); struct xinpgen { ksize_t xig_len; /* length of this structure */ u_int xig_count; /* number of PCBs at this time */ uint32_t _xig_spare32; inp_gen_t xig_gen; /* generation count at this time */ so_gen_t xig_sogen; /* socket generation count this time */ uint64_t _xig_spare64[4]; } __aligned(8); struct sockopt_parameters { struct in_conninfo sop_inc; uint64_t sop_id; int sop_level; int sop_optname; char sop_optval[]; }; #ifdef _KERNEL int sysctl_setsockopt(SYSCTL_HANDLER_ARGS, struct inpcbinfo *pcbinfo, int (*ctloutput_set)(struct inpcb *, struct sockopt *)); void in_pcbtoxinpcb(const struct inpcb *, struct xinpcb *); #endif #endif /* _SYS_SOCKETVAR_H_ */ #ifdef _KERNEL /* * Per-VNET pcb database for each high-level protocol (UDP, TCP, ...) in both * IPv4 and IPv6. * * The pcbs are protected with SMR section and thus all lists in inpcbinfo * are CK-lists. Locking is required to insert a pcb into database. Two * locks are provided: one for the hash and one for the global list of pcbs, * as well as overall count and generation count. * * Locking key: * * (c) Constant or nearly constant after initialisation * (e) Protected by SMR section * (g) Locked by ipi_lock * (h) Locked by ipi_hash_lock */ struct inpcbinfo { /* * Global lock protecting inpcb list modification */ struct mtx ipi_lock; struct inpcbhead ipi_listhead; /* (r:e/w:g) */ u_int ipi_count; /* (g) */ /* * Generation count -- incremented each time a connection is allocated * or freed. */ u_quad_t ipi_gencnt; /* (g) */ /* * Fields associated with port lookup and allocation. */ u_short ipi_lastport; /* (h) */ u_short ipi_lastlow; /* (h) */ u_short ipi_lasthi; /* (h) */ /* * UMA zone from which inpcbs are allocated for this protocol. */ uma_zone_t ipi_zone; /* (c) */ uma_zone_t ipi_portzone; /* (c) */ smr_t ipi_smr; /* (c) */ /* * Global hash of inpcbs, hashed by local and foreign addresses and * port numbers. The "exact" hash holds PCBs connected to a foreign * address, and "wild" holds the rest. */ struct mtx ipi_hash_lock; struct inpcbhead *ipi_hash_exact; /* (r:e/w:h) */ struct inpcbhead *ipi_hash_wild; /* (r:e/w:h) */ u_long ipi_hashmask; /* (c) */ /* * Global hash of inpcbs, hashed by only local port number. */ struct inpcbporthead *ipi_porthashbase; /* (h) */ u_long ipi_porthashmask; /* (h) */ /* * Load balance groups used for the SO_REUSEPORT_LB option, * hashed by local port. */ struct inpcblbgrouphead *ipi_lbgrouphashbase; /* (r:e/w:h) */ u_long ipi_lbgrouphashmask; /* (h) */ /* * Pointer to network stack instance */ struct vnet *ipi_vnet; /* (c) */ }; /* * Global allocation storage for each high-level protocol (UDP, TCP, ...). * Each corresponding per-VNET inpcbinfo points into this one. */ struct inpcbstorage { uma_zone_t ips_zone; uma_zone_t ips_portzone; uma_init ips_pcbinit; size_t ips_size; const char * ips_zone_name; const char * ips_portzone_name; const char * ips_infolock_name; const char * ips_hashlock_name; }; #define INPCBSTORAGE_DEFINE(prot, ppcb, lname, zname, iname, hname) \ static int \ prot##_inpcb_init(void *mem, int size __unused, int flags __unused) \ { \ struct inpcb *inp = mem; \ \ rw_init_flags(&inp->inp_lock, lname, RW_RECURSE | RW_DUPOK); \ return (0); \ } \ static struct inpcbstorage prot = { \ .ips_size = sizeof(struct ppcb), \ .ips_pcbinit = prot##_inpcb_init, \ .ips_zone_name = zname, \ .ips_portzone_name = zname " ports", \ .ips_infolock_name = iname, \ .ips_hashlock_name = hname, \ }; \ SYSINIT(prot##_inpcbstorage_init, SI_SUB_PROTO_DOMAIN, \ SI_ORDER_SECOND, in_pcbstorage_init, &prot); \ SYSUNINIT(prot##_inpcbstorage_uninit, SI_SUB_PROTO_DOMAIN, \ SI_ORDER_SECOND, in_pcbstorage_destroy, &prot) #define INP_LOCK_DESTROY(inp) rw_destroy(&(inp)->inp_lock) #define INP_RLOCK(inp) rw_rlock(&(inp)->inp_lock) #define INP_WLOCK(inp) rw_wlock(&(inp)->inp_lock) #define INP_TRY_RLOCK(inp) rw_try_rlock(&(inp)->inp_lock) #define INP_TRY_WLOCK(inp) rw_try_wlock(&(inp)->inp_lock) #define INP_RUNLOCK(inp) rw_runlock(&(inp)->inp_lock) #define INP_WUNLOCK(inp) rw_wunlock(&(inp)->inp_lock) #define INP_UNLOCK(inp) rw_unlock(&(inp)->inp_lock) #define INP_TRY_UPGRADE(inp) rw_try_upgrade(&(inp)->inp_lock) #define INP_DOWNGRADE(inp) rw_downgrade(&(inp)->inp_lock) #define INP_WLOCKED(inp) rw_wowned(&(inp)->inp_lock) #define INP_LOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_LOCKED) #define INP_RLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_RLOCKED) #define INP_WLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_WLOCKED) #define INP_UNLOCK_ASSERT(inp) rw_assert(&(inp)->inp_lock, RA_UNLOCKED) /* * These locking functions are for inpcb consumers outside of sys/netinet, * more specifically, they were added for the benefit of TOE drivers. The * macros are reserved for use by the stack. */ void inp_wlock(struct inpcb *); void inp_wunlock(struct inpcb *); void inp_rlock(struct inpcb *); void inp_runlock(struct inpcb *); #ifdef INVARIANT_SUPPORT void inp_lock_assert(struct inpcb *); void inp_unlock_assert(struct inpcb *); #else #define inp_lock_assert(inp) do {} while (0) #define inp_unlock_assert(inp) do {} while (0) #endif void inp_apply_all(struct inpcbinfo *, void (*func)(struct inpcb *, void *), void *arg); int inp_ip_tos_get(const struct inpcb *inp); void inp_ip_tos_set(struct inpcb *inp, int val); struct socket * inp_inpcbtosocket(struct inpcb *inp); struct tcpcb * inp_inpcbtotcpcb(struct inpcb *inp); void inp_4tuple_get(struct inpcb *inp, uint32_t *laddr, uint16_t *lp, uint32_t *faddr, uint16_t *fp); #endif /* _KERNEL */ #define INP_INFO_WLOCK(ipi) mtx_lock(&(ipi)->ipi_lock) #define INP_INFO_WLOCKED(ipi) mtx_owned(&(ipi)->ipi_lock) #define INP_INFO_WUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_lock) #define INP_INFO_LOCK_ASSERT(ipi) MPASS(SMR_ENTERED((ipi)->ipi_smr) || \ mtx_owned(&(ipi)->ipi_lock)) #define INP_INFO_WLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_lock, MA_OWNED) #define INP_INFO_WUNLOCK_ASSERT(ipi) \ mtx_assert(&(ipi)->ipi_lock, MA_NOTOWNED) #define INP_HASH_WLOCK(ipi) mtx_lock(&(ipi)->ipi_hash_lock) #define INP_HASH_WUNLOCK(ipi) mtx_unlock(&(ipi)->ipi_hash_lock) #define INP_HASH_LOCK_ASSERT(ipi) MPASS(SMR_ENTERED((ipi)->ipi_smr) || \ mtx_owned(&(ipi)->ipi_hash_lock)) #define INP_HASH_WLOCK_ASSERT(ipi) mtx_assert(&(ipi)->ipi_hash_lock, \ MA_OWNED) /* * Wildcard matching hash is not just a microoptimisation! The hash for * wildcard IPv4 and wildcard IPv6 must be the same, otherwise AF_INET6 * wildcard bound pcb won't be able to receive AF_INET connections, while: * jenkins_hash(&zeroes, 1, s) != jenkins_hash(&zeroes, 4, s) * See also comment above struct in_addr_4in6. */ #define IN_ADDR_JHASH32(addr) \ ((addr)->s_addr == INADDR_ANY ? V_in_pcbhashseed : \ jenkins_hash32((&(addr)->s_addr), 1, V_in_pcbhashseed)) #define IN6_ADDR_JHASH32(addr) \ (memcmp((addr), &in6addr_any, sizeof(in6addr_any)) == 0 ? \ V_in_pcbhashseed : \ jenkins_hash32((addr)->__u6_addr.__u6_addr32, \ nitems((addr)->__u6_addr.__u6_addr32), V_in_pcbhashseed)) #define INP_PCBHASH(faddr, lport, fport, mask) \ ((IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask)) #define INP6_PCBHASH(faddr, lport, fport, mask) \ ((IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) & (mask)) #define INP_PCBHASH_WILD(lport, mask) \ ((V_in_pcbhashseed ^ ntohs(lport)) & (mask)) #define INP_PCBLBGROUP_PKTHASH(faddr, lport, fport) \ (IN_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) #define INP6_PCBLBGROUP_PKTHASH(faddr, lport, fport) \ (IN6_ADDR_JHASH32(faddr) ^ ntohs((lport) ^ (fport))) #define INP_PCBPORTHASH(lport, mask) (ntohs((lport)) & (mask)) /* * Flags for inp_vflags -- historically version flags only */ #define INP_IPV4 0x1 #define INP_IPV6 0x2 #define INP_IPV6PROTO 0x4 /* opened under IPv6 protocol */ /* * Flags for inp_flags. */ #define INP_RECVOPTS 0x00000001 /* receive incoming IP options */ #define INP_RECVRETOPTS 0x00000002 /* receive IP options for reply */ #define INP_RECVDSTADDR 0x00000004 /* receive IP dst address */ #define INP_HDRINCL 0x00000008 /* user supplies entire IP header */ #define INP_HIGHPORT 0x00000010 /* user wants "high" port binding */ #define INP_LOWPORT 0x00000020 /* user wants "low" port binding */ #define INP_ANONPORT 0x00000040 /* read by netstat(1) */ #define INP_RECVIF 0x00000080 /* receive incoming interface */ #define INP_MTUDISC 0x00000100 /* user can do MTU discovery */ /* INP_FREED 0x00000200 private to in_pcb.c */ #define INP_RECVTTL 0x00000400 /* receive incoming IP TTL */ #define INP_DONTFRAG 0x00000800 /* don't fragment packet */ #define INP_BINDANY 0x00001000 /* allow bind to any address */ #define INP_INHASHLIST 0x00002000 /* in_pcbinshash() has been called */ #define INP_RECVTOS 0x00004000 /* receive incoming IP TOS */ #define IN6P_IPV6_V6ONLY 0x00008000 /* restrict AF_INET6 socket for v6 */ #define IN6P_PKTINFO 0x00010000 /* receive IP6 dst and I/F */ #define IN6P_HOPLIMIT 0x00020000 /* receive hoplimit */ #define IN6P_HOPOPTS 0x00040000 /* receive hop-by-hop options */ #define IN6P_DSTOPTS 0x00080000 /* receive dst options after rthdr */ #define IN6P_RTHDR 0x00100000 /* receive routing header */ #define IN6P_RTHDRDSTOPTS 0x00200000 /* receive dstoptions before rthdr */ #define IN6P_TCLASS 0x00400000 /* receive traffic class value */ #define IN6P_AUTOFLOWLABEL 0x00800000 /* attach flowlabel automatically */ /* INP_INLBGROUP 0x01000000 private to in_pcb.c */ #define INP_ONESBCAST 0x02000000 /* send all-ones broadcast */ #define INP_DROPPED 0x04000000 /* protocol drop flag */ #define INP_SOCKREF 0x08000000 /* strong socket reference */ #define INP_RESERVED_0 0x10000000 /* reserved field */ #define INP_BOUNDFIB 0x20000000 /* Bound to a specific FIB. */ #define IN6P_RFC2292 0x40000000 /* used RFC2292 API on the socket */ #define IN6P_MTU 0x80000000 /* receive path MTU */ #define INP_CONTROLOPTS (INP_RECVOPTS|INP_RECVRETOPTS|INP_RECVDSTADDR|\ INP_RECVIF|INP_RECVTTL|INP_RECVTOS|\ IN6P_PKTINFO|IN6P_HOPLIMIT|IN6P_HOPOPTS|\ IN6P_DSTOPTS|IN6P_RTHDR|IN6P_RTHDRDSTOPTS|\ IN6P_TCLASS|IN6P_AUTOFLOWLABEL|IN6P_RFC2292|\ IN6P_MTU) /* * Flags for inp_flags2. */ /* 0x00000001 */ /* 0x00000002 */ /* 0x00000004 */ /* 0x00000008 */ /* 0x00000010 */ /* 0x00000020 */ /* 0x00000040 */ /* 0x00000080 */ #define INP_RECVFLOWID 0x00000100 /* populate recv datagram with flow info */ #define INP_RECVRSSBUCKETID 0x00000200 /* populate recv datagram with bucket id */ #define INP_RATE_LIMIT_CHANGED 0x00000400 /* rate limit needs attention */ #define INP_ORIGDSTADDR 0x00000800 /* receive IP dst address/port */ /* 0x00001000 */ /* 0x00002000 */ /* 0x00004000 */ /* 0x00008000 */ /* 0x00010000 */ #define INP_2PCP_SET 0x00020000 /* If the Eth PCP should be set explicitly */ #define INP_2PCP_BIT0 0x00040000 /* Eth PCP Bit 0 */ #define INP_2PCP_BIT1 0x00080000 /* Eth PCP Bit 1 */ #define INP_2PCP_BIT2 0x00100000 /* Eth PCP Bit 2 */ #define INP_2PCP_BASE INP_2PCP_BIT0 #define INP_2PCP_MASK (INP_2PCP_BIT0 | INP_2PCP_BIT1 | INP_2PCP_BIT2) #define INP_2PCP_SHIFT 18 /* shift PCP field in/out of inp_flags2 */ /* * Flags passed to in_pcblookup*(), inp_smr_lock() and inp_next(). */ typedef enum { INPLOOKUP_WILDCARD = 0x00000001, /* Allow wildcard sockets. */ INPLOOKUP_RLOCKPCB = 0x00000002, /* Return inpcb read-locked. */ INPLOOKUP_WLOCKPCB = 0x00000004, /* Return inpcb write-locked. */ INPLOOKUP_FIB = 0x00000008, /* inp must be from same FIB. */ } inp_lookup_t; #define INPLOOKUP_MASK (INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB | \ INPLOOKUP_WLOCKPCB | INPLOOKUP_FIB) #define INPLOOKUP_LOCKMASK (INPLOOKUP_RLOCKPCB | INPLOOKUP_WLOCKPCB) #define sotoinpcb(so) ((struct inpcb *)(so)->so_pcb) #define INP_SOCKAF(so) so->so_proto->pr_domain->dom_family #define INP_CHECK_SOCKAF(so, af) (INP_SOCKAF(so) == af) #ifdef _KERNEL VNET_DECLARE(int, ipport_reservedhigh); VNET_DECLARE(int, ipport_reservedlow); VNET_DECLARE(int, ipport_lowfirstauto); VNET_DECLARE(int, ipport_lowlastauto); VNET_DECLARE(int, ipport_firstauto); VNET_DECLARE(int, ipport_lastauto); VNET_DECLARE(int, ipport_hifirstauto); VNET_DECLARE(int, ipport_hilastauto); VNET_DECLARE(int, ipport_randomized); #define V_ipport_reservedhigh VNET(ipport_reservedhigh) #define V_ipport_reservedlow VNET(ipport_reservedlow) #define V_ipport_lowfirstauto VNET(ipport_lowfirstauto) #define V_ipport_lowlastauto VNET(ipport_lowlastauto) #define V_ipport_firstauto VNET(ipport_firstauto) #define V_ipport_lastauto VNET(ipport_lastauto) #define V_ipport_hifirstauto VNET(ipport_hifirstauto) #define V_ipport_hilastauto VNET(ipport_hilastauto) #define V_ipport_randomized VNET(ipport_randomized) void in_pcbinfo_init(struct inpcbinfo *, struct inpcbstorage *, u_int, u_int); void in_pcbinfo_destroy(struct inpcbinfo *); void in_pcbstorage_init(void *); void in_pcbstorage_destroy(void *); void in_pcbpurgeif0(struct inpcbinfo *, struct ifnet *); int in_pcballoc(struct socket *, struct inpcbinfo *); #define INPBIND_FIB 0x0001 /* bind to the PCB's FIB only */ int in_pcbbind(struct inpcb *, struct sockaddr_in *, int, struct ucred *); int in_pcbbind_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *, u_short *, int, struct ucred *); int in_pcbconnect(struct inpcb *, struct sockaddr_in *, struct ucred *, bool); int in_pcbconnect_setup(struct inpcb *, struct sockaddr_in *, in_addr_t *, u_short *, in_addr_t *, u_short *, struct ucred *); void in_pcbdisconnect(struct inpcb *); void in_pcbdrop(struct inpcb *); void in_pcbfree(struct inpcb *); int in_pcbinshash(struct inpcb *); int in_pcbladdr(struct inpcb *, struct in_addr *, struct in_addr *, struct ucred *); int in_pcblbgroup_numa(struct inpcb *, int arg); struct inpcb * in_pcblookup(struct inpcbinfo *, struct in_addr, u_int, struct in_addr, u_int, int, struct ifnet *); struct inpcb * in_pcblookup_mbuf(struct inpcbinfo *, struct in_addr, u_int, struct in_addr, u_int, int, struct ifnet *, struct mbuf *); void in_pcbnotifyall(struct inpcbinfo *pcbinfo, struct in_addr, int, struct inpcb *(*)(struct inpcb *, int)); void in_pcbref(struct inpcb *); void in_pcbrehash(struct inpcb *); void in_pcbremhash_locked(struct inpcb *); bool in_pcbrele(struct inpcb *, inp_lookup_t); bool in_pcbrele_rlocked(struct inpcb *); bool in_pcbrele_wlocked(struct inpcb *); typedef bool inp_match_t(const struct inpcb *, void *); struct inpcb_iterator { const struct inpcbinfo *ipi; struct inpcb *inp; inp_match_t *match; void *ctx; int hash; #define INP_ALL_LIST -1 const inp_lookup_t lock; }; /* Note: sparse initializers guarantee .inp = NULL. */ #define INP_ITERATOR(_ipi, _lock, _match, _ctx) \ { \ .ipi = (_ipi), \ .lock = (_lock), \ .hash = INP_ALL_LIST, \ .match = (_match), \ .ctx = (_ctx), \ } #define INP_ALL_ITERATOR(_ipi, _lock) \ { \ .ipi = (_ipi), \ .lock = (_lock), \ .hash = INP_ALL_LIST, \ } struct inpcb *inp_next(struct inpcb_iterator *); void in_losing(struct inpcb *); void in_pcbsetsolabel(struct socket *so); int in_getpeeraddr(struct socket *so, struct sockaddr **nam); int in_getsockaddr(struct socket *so, struct sockaddr **nam); struct sockaddr * in_sockaddr(in_port_t port, struct in_addr *addr); void in_pcbsosetlabel(struct socket *so); #ifdef RATELIMIT int in_pcboutput_txrtlmt_locked(struct inpcb *, struct ifnet *, struct mbuf *, uint32_t); int in_pcbattach_txrtlmt(struct inpcb *, struct ifnet *, uint32_t, uint32_t, uint32_t, struct m_snd_tag **); void in_pcbdetach_txrtlmt(struct inpcb *); void in_pcbdetach_tag(struct m_snd_tag *); int in_pcbmodify_txrtlmt(struct inpcb *, uint32_t); int in_pcbquery_txrtlmt(struct inpcb *, uint32_t *); int in_pcbquery_txrlevel(struct inpcb *, uint32_t *); void in_pcboutput_txrtlmt(struct inpcb *, struct ifnet *, struct mbuf *); void in_pcboutput_eagain(struct inpcb *); #endif +#ifdef DDB +void db_print_inpcb(struct inpcb *, const char *, int); +#endif #endif /* _KERNEL */ #endif /* !_NETINET_IN_PCB_H_ */ diff --git a/sys/netinet/tcp_usrreq.c b/sys/netinet/tcp_usrreq.c index daa04845ab79..38c640569f5b 100644 --- a/sys/netinet/tcp_usrreq.c +++ b/sys/netinet/tcp_usrreq.c @@ -1,3293 +1,3300 @@ /*- * SPDX-License-Identifier: BSD-3-Clause * * Copyright (c) 1982, 1986, 1988, 1993 * The Regents of the University of California. * Copyright (c) 2006-2007 Robert N. M. Watson * Copyright (c) 2010-2011 Juniper Networks, Inc. * All rights reserved. * * Portions of this software were developed by Robert N. M. Watson under * contract to Juniper Networks, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * 3. Neither the name of the University nor the names of its contributors * may be used to endorse or promote products derived from this software * without specific prior written permission. * * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. * * From: @(#)tcp_usrreq.c 8.2 (Berkeley) 1/3/94 */ #include #include "opt_ddb.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_ipsec.h" #include "opt_kern_tls.h" #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #endif /* INET6 */ #include #include #include #include #include #include #ifdef DDB #include #endif #include #include #include #include #include #include #include #include #include #include #include #ifdef INET6 #include #include #include #include #endif #include #include #include #include #include #include #include #include #include #include #ifdef TCPPCAP #include #endif #ifdef TCP_OFFLOAD #include #endif #include #include #include #include #include #include #include /* * TCP protocol interface to socket abstraction. */ #ifdef INET static int tcp_connect(struct tcpcb *, struct sockaddr_in *, struct thread *td); #endif /* INET */ #ifdef INET6 static int tcp6_connect(struct tcpcb *, struct sockaddr_in6 *, struct thread *td); #endif /* INET6 */ static void tcp_disconnect(struct tcpcb *); static void tcp_usrclosed(struct tcpcb *); static void tcp_fill_info(const struct tcpcb *, struct tcp_info *); static int tcp_pru_options_support(struct tcpcb *tp, int flags); static void tcp_bblog_pru(struct tcpcb *tp, uint32_t pru, int error) { struct tcp_log_buffer *lgb; KASSERT(tp != NULL, ("tcp_bblog_pru: tp == NULL")); INP_WLOCK_ASSERT(tptoinpcb(tp)); if (tcp_bblogging_on(tp)) { lgb = tcp_log_event(tp, NULL, NULL, NULL, TCP_LOG_PRU, error, 0, NULL, false, NULL, NULL, 0, NULL); } else { lgb = NULL; } if (lgb != NULL) { if (error >= 0) { lgb->tlb_errno = (uint32_t)error; } lgb->tlb_flex1 = pru; } } /* * TCP attaches to socket via pru_attach(), reserving space, * and an internet control block. */ static int tcp_usr_attach(struct socket *so, int proto, struct thread *td) { struct inpcb *inp; struct tcpcb *tp = NULL; int error; inp = sotoinpcb(so); KASSERT(inp == NULL, ("tcp_usr_attach: inp != NULL")); error = soreserve(so, V_tcp_sendspace, V_tcp_recvspace); if (error) goto out; so->so_rcv.sb_flags |= SB_AUTOSIZE; so->so_snd.sb_flags |= SB_AUTOSIZE; error = in_pcballoc(so, &V_tcbinfo); if (error) goto out; inp = sotoinpcb(so); tp = tcp_newtcpcb(inp, NULL); if (tp == NULL) { error = ENOBUFS; in_pcbfree(inp); goto out; } tp->t_state = TCPS_CLOSED; tcp_bblog_pru(tp, PRU_ATTACH, error); INP_WUNLOCK(inp); TCPSTATES_INC(TCPS_CLOSED); out: TCP_PROBE2(debug__user, tp, PRU_ATTACH); return (error); } /* * tcp_usr_detach is called when the socket layer loses its final reference * to the socket, be it a file descriptor reference, a reference from TCP, * etc. At this point, there is only one case in which we will keep around * inpcb state: time wait. */ static void tcp_usr_detach(struct socket *so) { struct inpcb *inp; struct tcpcb *tp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("%s: inp == NULL", __func__)); INP_WLOCK(inp); KASSERT(so->so_pcb == inp && inp->inp_socket == so, ("%s: socket %p inp %p mismatch", __func__, so, inp)); tp = intotcpcb(inp); KASSERT(inp->inp_flags & INP_DROPPED || tp->t_state < TCPS_SYN_SENT, ("%s: inp %p not dropped or embryonic", __func__, inp)); tcp_discardcb(tp); in_pcbfree(inp); } #ifdef INET /* * Give the socket an address. */ static int tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp; struct sockaddr_in *sinp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_bind: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (EINVAL); } tp = intotcpcb(inp); sinp = (struct sockaddr_in *)nam; if (nam->sa_family != AF_INET) { /* * Preserve compatibility with old programs. */ if (nam->sa_family != AF_UNSPEC || nam->sa_len < offsetof(struct sockaddr_in, sin_zero) || sinp->sin_addr.s_addr != INADDR_ANY) { error = EAFNOSUPPORT; goto out; } nam->sa_family = AF_INET; } if (nam->sa_len != sizeof(*sinp)) { error = EINVAL; goto out; } /* * Must check for multicast addresses and disallow binding * to them. */ if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { error = EAFNOSUPPORT; goto out; } INP_HASH_WLOCK(&V_tcbinfo); error = in_pcbbind(inp, sinp, V_tcp_bind_all_fibs ? 0 : INPBIND_FIB, td->td_ucred); INP_HASH_WUNLOCK(&V_tcbinfo); out: tcp_bblog_pru(tp, PRU_BIND, error); TCP_PROBE2(debug__user, tp, PRU_BIND); INP_WUNLOCK(inp); return (error); } #endif /* INET */ #ifdef INET6 static int tcp6_usr_bind(struct socket *so, struct sockaddr *nam, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp; struct sockaddr_in6 *sin6; u_char vflagsav; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_bind: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (EINVAL); } tp = intotcpcb(inp); vflagsav = inp->inp_vflag; sin6 = (struct sockaddr_in6 *)nam; if (nam->sa_family != AF_INET6) { error = EAFNOSUPPORT; goto out; } if (nam->sa_len != sizeof(*sin6)) { error = EINVAL; goto out; } /* * Must check for multicast addresses and disallow binding * to them. */ if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { error = EAFNOSUPPORT; goto out; } INP_HASH_WLOCK(&V_tcbinfo); inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; #ifdef INET if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) { if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) inp->inp_vflag |= INP_IPV4; else if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { struct sockaddr_in sin; in6_sin6_2_sin(&sin, sin6); if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { error = EAFNOSUPPORT; INP_HASH_WUNLOCK(&V_tcbinfo); goto out; } inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; error = in_pcbbind(inp, &sin, 0, td->td_ucred); INP_HASH_WUNLOCK(&V_tcbinfo); goto out; } } #endif error = in6_pcbbind(inp, sin6, V_tcp_bind_all_fibs ? 0 : INPBIND_FIB, td->td_ucred); INP_HASH_WUNLOCK(&V_tcbinfo); out: if (error != 0) inp->inp_vflag = vflagsav; tcp_bblog_pru(tp, PRU_BIND, error); TCP_PROBE2(debug__user, tp, PRU_BIND); INP_WUNLOCK(inp); return (error); } #endif /* INET6 */ #ifdef INET /* * Prepare to accept connections. */ static int tcp_usr_listen(struct socket *so, int backlog, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_listen: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (EINVAL); } tp = intotcpcb(inp); SOCK_LOCK(so); error = solisten_proto_check(so); if (error != 0) { SOCK_UNLOCK(so); goto out; } if (inp->inp_lport == 0) { INP_HASH_WLOCK(&V_tcbinfo); error = in_pcbbind(inp, NULL, V_tcp_bind_all_fibs ? 0 : INPBIND_FIB, td->td_ucred); INP_HASH_WUNLOCK(&V_tcbinfo); } if (error == 0) { tcp_state_change(tp, TCPS_LISTEN); solisten_proto(so, backlog); #ifdef TCP_OFFLOAD if ((so->so_options & SO_NO_OFFLOAD) == 0) tcp_offload_listen_start(tp); #endif } else { solisten_proto_abort(so); } SOCK_UNLOCK(so); if (IS_FASTOPEN(tp->t_flags)) tp->t_tfo_pending = tcp_fastopen_alloc_counter(); out: tcp_bblog_pru(tp, PRU_LISTEN, error); TCP_PROBE2(debug__user, tp, PRU_LISTEN); INP_WUNLOCK(inp); return (error); } #endif /* INET */ #ifdef INET6 static int tcp6_usr_listen(struct socket *so, int backlog, struct thread *td) { int error = 0; struct inpcb *inp; struct tcpcb *tp; u_char vflagsav; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_listen: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (EINVAL); } tp = intotcpcb(inp); vflagsav = inp->inp_vflag; SOCK_LOCK(so); error = solisten_proto_check(so); if (error != 0) { SOCK_UNLOCK(so); goto out; } INP_HASH_WLOCK(&V_tcbinfo); if (inp->inp_lport == 0) { inp->inp_vflag &= ~INP_IPV4; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) == 0) inp->inp_vflag |= INP_IPV4; error = in6_pcbbind(inp, NULL, V_tcp_bind_all_fibs ? 0 : INPBIND_FIB, td->td_ucred); } INP_HASH_WUNLOCK(&V_tcbinfo); if (error == 0) { tcp_state_change(tp, TCPS_LISTEN); solisten_proto(so, backlog); #ifdef TCP_OFFLOAD if ((so->so_options & SO_NO_OFFLOAD) == 0) tcp_offload_listen_start(tp); #endif } else { solisten_proto_abort(so); } SOCK_UNLOCK(so); if (IS_FASTOPEN(tp->t_flags)) tp->t_tfo_pending = tcp_fastopen_alloc_counter(); if (error != 0) inp->inp_vflag = vflagsav; out: tcp_bblog_pru(tp, PRU_LISTEN, error); TCP_PROBE2(debug__user, tp, PRU_LISTEN); INP_WUNLOCK(inp); return (error); } #endif /* INET6 */ #ifdef INET /* * Initiate connection to peer. * Create a template for use in transmissions on this connection. * Enter SYN_SENT state, and mark socket as connecting. * Start keep-alive timer, and seed output sequence space. * Send initial segment on connection. */ static int tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct epoch_tracker et; int error = 0; struct inpcb *inp; struct tcpcb *tp; struct sockaddr_in *sinp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_connect: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (ECONNREFUSED); } tp = intotcpcb(inp); sinp = (struct sockaddr_in *)nam; if (nam->sa_family != AF_INET) { error = EAFNOSUPPORT; goto out; } if (nam->sa_len != sizeof (*sinp)) { error = EINVAL; goto out; } /* * Must disallow TCP ``connections'' to multicast addresses. */ if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { error = EAFNOSUPPORT; goto out; } if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) { error = EACCES; goto out; } if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr)) != 0) goto out; if (SOLISTENING(so)) { error = EOPNOTSUPP; goto out; } NET_EPOCH_ENTER(et); if ((error = tcp_connect(tp, sinp, td)) != 0) goto out_in_epoch; #ifdef TCP_OFFLOAD if (registered_toedevs > 0 && (so->so_options & SO_NO_OFFLOAD) == 0 && (error = tcp_offload_connect(so, nam)) == 0) goto out_in_epoch; #endif tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); error = tcp_output(tp); KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); out_in_epoch: NET_EPOCH_EXIT(et); out: tcp_bblog_pru(tp, PRU_CONNECT, error); TCP_PROBE2(debug__user, tp, PRU_CONNECT); INP_WUNLOCK(inp); return (error); } #endif /* INET */ #ifdef INET6 static int tcp6_usr_connect(struct socket *so, struct sockaddr *nam, struct thread *td) { struct epoch_tracker et; int error = 0; struct inpcb *inp; struct tcpcb *tp; struct sockaddr_in6 *sin6; u_int8_t incflagsav; u_char vflagsav; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_connect: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (ECONNREFUSED); } tp = intotcpcb(inp); vflagsav = inp->inp_vflag; incflagsav = inp->inp_inc.inc_flags; sin6 = (struct sockaddr_in6 *)nam; if (nam->sa_family != AF_INET6) { error = EAFNOSUPPORT; goto out; } if (nam->sa_len != sizeof (*sin6)) { error = EINVAL; goto out; } /* * Must disallow TCP ``connections'' to multicast addresses. */ if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { error = EAFNOSUPPORT; goto out; } if (SOLISTENING(so)) { error = EINVAL; goto out; } #ifdef INET /* * XXXRW: Some confusion: V4/V6 flags relate to binding, and * therefore probably require the hash lock, which isn't held here. * Is this a significant problem? */ if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { struct sockaddr_in sin; if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { error = EINVAL; goto out; } if ((inp->inp_vflag & INP_IPV4) == 0) { error = EAFNOSUPPORT; goto out; } in6_sin6_2_sin(&sin, sin6); if (IN_MULTICAST(ntohl(sin.sin_addr.s_addr))) { error = EAFNOSUPPORT; goto out; } if (ntohl(sin.sin_addr.s_addr) == INADDR_BROADCAST) { error = EACCES; goto out; } if ((error = prison_remote_ip4(td->td_ucred, &sin.sin_addr)) != 0) goto out; inp->inp_vflag |= INP_IPV4; inp->inp_vflag &= ~INP_IPV6; NET_EPOCH_ENTER(et); if ((error = tcp_connect(tp, &sin, td)) != 0) goto out_in_epoch; #ifdef TCP_OFFLOAD if (registered_toedevs > 0 && (so->so_options & SO_NO_OFFLOAD) == 0 && (error = tcp_offload_connect(so, nam)) == 0) goto out_in_epoch; #endif error = tcp_output(tp); goto out_in_epoch; } else { if ((inp->inp_vflag & INP_IPV6) == 0) { error = EAFNOSUPPORT; goto out; } } #endif if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr)) != 0) goto out; inp->inp_vflag &= ~INP_IPV4; inp->inp_vflag |= INP_IPV6; inp->inp_inc.inc_flags |= INC_ISIPV6; NET_EPOCH_ENTER(et); if ((error = tcp6_connect(tp, sin6, td)) != 0) goto out_in_epoch; #ifdef TCP_OFFLOAD if (registered_toedevs > 0 && (so->so_options & SO_NO_OFFLOAD) == 0 && (error = tcp_offload_connect(so, nam)) == 0) goto out_in_epoch; #endif tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); error = tcp_output(tp); out_in_epoch: NET_EPOCH_EXIT(et); out: KASSERT(error >= 0, ("TCP stack %s requested tcp_drop(%p) at connect()" ", error code %d", tp->t_fb->tfb_tcp_block_name, tp, -error)); /* * If the implicit bind in the connect call fails, restore * the flags we modified. */ if (error != 0 && inp->inp_lport == 0) { inp->inp_vflag = vflagsav; inp->inp_inc.inc_flags = incflagsav; } tcp_bblog_pru(tp, PRU_CONNECT, error); TCP_PROBE2(debug__user, tp, PRU_CONNECT); INP_WUNLOCK(inp); return (error); } #endif /* INET6 */ /* * Initiate disconnect from peer. * If connection never passed embryonic stage, just drop; * else if don't need to let data drain, then can just drop anyways, * else have to begin TCP shutdown process: mark socket disconnecting, * drain unread data, state switch to reflect user close, and * send segment (e.g. FIN) to peer. Socket will be really disconnected * when peer sends FIN and acks ours. * * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB. */ static int tcp_usr_disconnect(struct socket *so) { struct inpcb *inp; struct tcpcb *tp = NULL; struct epoch_tracker et; int error = 0; NET_EPOCH_ENTER(et); inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_disconnect: inp == NULL")); INP_WLOCK(inp); tp = intotcpcb(inp); if (tp->t_state == TCPS_TIME_WAIT) goto out; tcp_disconnect(tp); out: tcp_bblog_pru(tp, PRU_DISCONNECT, error); TCP_PROBE2(debug__user, tp, PRU_DISCONNECT); INP_WUNLOCK(inp); NET_EPOCH_EXIT(et); return (error); } #ifdef INET /* * Accept a connection. Essentially all the work is done at higher levels; * just return the address of the peer, storing through addr. */ static int tcp_usr_accept(struct socket *so, struct sockaddr **nam) { int error = 0; struct inpcb *inp; struct tcpcb *tp; struct in_addr addr; in_port_t port = 0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_accept: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (ECONNABORTED); } tp = intotcpcb(inp); if (so->so_state & SS_ISDISCONNECTED) { error = ECONNABORTED; goto out; } /* * We inline in_getpeeraddr and COMMON_END here, so that we can * copy the data of interest and defer the malloc until after we * release the lock. */ port = inp->inp_fport; addr = inp->inp_faddr; out: tcp_bblog_pru(tp, PRU_ACCEPT, error); TCP_PROBE2(debug__user, tp, PRU_ACCEPT); INP_WUNLOCK(inp); if (error == 0) *nam = in_sockaddr(port, &addr); return error; } #endif /* INET */ #ifdef INET6 static int tcp6_usr_accept(struct socket *so, struct sockaddr **nam) { struct inpcb *inp; int error = 0; struct tcpcb *tp; struct in_addr addr; struct in6_addr addr6; struct epoch_tracker et; in_port_t port = 0; int v4 = 0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp6_usr_accept: inp == NULL")); NET_EPOCH_ENTER(et); /* XXXMT Why is this needed? */ INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); NET_EPOCH_EXIT(et); return (ECONNABORTED); } tp = intotcpcb(inp); if (so->so_state & SS_ISDISCONNECTED) { error = ECONNABORTED; goto out; } /* * We inline in6_mapped_peeraddr and COMMON_END here, so that we can * copy the data of interest and defer the malloc until after we * release the lock. */ if (inp->inp_vflag & INP_IPV4) { v4 = 1; port = inp->inp_fport; addr = inp->inp_faddr; } else { port = inp->inp_fport; addr6 = inp->in6p_faddr; } out: tcp_bblog_pru(tp, PRU_ACCEPT, error); TCP_PROBE2(debug__user, tp, PRU_ACCEPT); INP_WUNLOCK(inp); NET_EPOCH_EXIT(et); if (error == 0) { if (v4) *nam = in6_v4mapsin6_sockaddr(port, &addr); else *nam = in6_sockaddr(port, &addr6); } return error; } #endif /* INET6 */ /* * Mark the connection as being incapable of further output. */ static int tcp_usr_shutdown(struct socket *so) { int error = 0; struct inpcb *inp; struct tcpcb *tp; struct epoch_tracker et; inp = sotoinpcb(so); KASSERT(inp != NULL, ("inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (ECONNRESET); } tp = intotcpcb(inp); NET_EPOCH_ENTER(et); socantsendmore(so); tcp_usrclosed(tp); if (!(inp->inp_flags & INP_DROPPED)) error = tcp_output_nodrop(tp); tcp_bblog_pru(tp, PRU_SHUTDOWN, error); TCP_PROBE2(debug__user, tp, PRU_SHUTDOWN); error = tcp_unlock_or_drop(tp, error); NET_EPOCH_EXIT(et); return (error); } /* * After a receive, possibly send window update to peer. */ static int tcp_usr_rcvd(struct socket *so, int flags) { struct epoch_tracker et; struct inpcb *inp; struct tcpcb *tp; int outrv = 0, error = 0; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_rcvd: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (ECONNRESET); } tp = intotcpcb(inp); NET_EPOCH_ENTER(et); /* * For passively-created TFO connections, don't attempt a window * update while still in SYN_RECEIVED as this may trigger an early * SYN|ACK. It is preferable to have the SYN|ACK be sent along with * application response data, or failing that, when the DELACK timer * expires. */ if (IS_FASTOPEN(tp->t_flags) && (tp->t_state == TCPS_SYN_RECEIVED)) goto out; #ifdef TCP_OFFLOAD if (tp->t_flags & TF_TOE) tcp_offload_rcvd(tp); else #endif outrv = tcp_output_nodrop(tp); out: tcp_bblog_pru(tp, PRU_RCVD, error); TCP_PROBE2(debug__user, tp, PRU_RCVD); (void) tcp_unlock_or_drop(tp, outrv); NET_EPOCH_EXIT(et); return (error); } /* * Do a send by putting data in output queue and updating urgent * marker if URG set. Possibly send more data. Unlike the other * pru_*() routines, the mbuf chains are our responsibility. We * must either enqueue them or free them. The other pru_* routines * generally are caller-frees. */ static int tcp_usr_send(struct socket *so, int flags, struct mbuf *m, struct sockaddr *nam, struct mbuf *control, struct thread *td) { struct epoch_tracker et; int error = 0; struct inpcb *inp; struct tcpcb *tp; #ifdef INET #ifdef INET6 struct sockaddr_in sin; #endif struct sockaddr_in *sinp; #endif #ifdef INET6 struct sockaddr_in6 *sin6; int isipv6; #endif u_int8_t incflagsav; u_char vflagsav; bool restoreflags; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { if (m != NULL && (flags & PRUS_NOTREADY) == 0) m_freem(m); INP_WUNLOCK(inp); return (ECONNRESET); } tp = intotcpcb(inp); vflagsav = inp->inp_vflag; incflagsav = inp->inp_inc.inc_flags; restoreflags = false; NET_EPOCH_ENTER(et); if (control != NULL) { /* TCP doesn't do control messages (rights, creds, etc) */ if (control->m_len > 0) { m_freem(control); error = EINVAL; goto out; } m_freem(control); /* empty control, just free it */ } if ((flags & PRUS_OOB) != 0 && (error = tcp_pru_options_support(tp, PRUS_OOB)) != 0) goto out; if (nam != NULL && tp->t_state < TCPS_SYN_SENT) { if (tp->t_state == TCPS_LISTEN) { error = EINVAL; goto out; } switch (nam->sa_family) { #ifdef INET case AF_INET: sinp = (struct sockaddr_in *)nam; if (sinp->sin_len != sizeof(struct sockaddr_in)) { error = EINVAL; goto out; } if ((inp->inp_vflag & INP_IPV6) != 0) { error = EAFNOSUPPORT; goto out; } if (IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) { error = EAFNOSUPPORT; goto out; } if (ntohl(sinp->sin_addr.s_addr) == INADDR_BROADCAST) { error = EACCES; goto out; } if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr))) goto out; #ifdef INET6 isipv6 = 0; #endif break; #endif /* INET */ #ifdef INET6 case AF_INET6: sin6 = (struct sockaddr_in6 *)nam; if (sin6->sin6_len != sizeof(*sin6)) { error = EINVAL; goto out; } if ((inp->inp_vflag & INP_IPV6PROTO) == 0) { error = EAFNOSUPPORT; goto out; } if (IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr)) { error = EAFNOSUPPORT; goto out; } if (IN6_IS_ADDR_V4MAPPED(&sin6->sin6_addr)) { #ifdef INET if ((inp->inp_flags & IN6P_IPV6_V6ONLY) != 0) { error = EINVAL; goto out; } if ((inp->inp_vflag & INP_IPV4) == 0) { error = EAFNOSUPPORT; goto out; } restoreflags = true; inp->inp_vflag &= ~INP_IPV6; sinp = &sin; in6_sin6_2_sin(sinp, sin6); if (IN_MULTICAST( ntohl(sinp->sin_addr.s_addr))) { error = EAFNOSUPPORT; goto out; } if ((error = prison_remote_ip4(td->td_ucred, &sinp->sin_addr))) goto out; isipv6 = 0; #else /* !INET */ error = EAFNOSUPPORT; goto out; #endif /* INET */ } else { if ((inp->inp_vflag & INP_IPV6) == 0) { error = EAFNOSUPPORT; goto out; } restoreflags = true; inp->inp_vflag &= ~INP_IPV4; inp->inp_inc.inc_flags |= INC_ISIPV6; if ((error = prison_remote_ip6(td->td_ucred, &sin6->sin6_addr))) goto out; isipv6 = 1; } break; #endif /* INET6 */ default: error = EAFNOSUPPORT; goto out; } } if (!(flags & PRUS_OOB)) { if (tp->t_acktime == 0) tp->t_acktime = ticks; sbappendstream(&so->so_snd, m, flags); m = NULL; if (nam && tp->t_state < TCPS_SYN_SENT) { KASSERT(tp->t_state == TCPS_CLOSED, ("%s: tp %p is listening", __func__, tp)); /* * Do implied connect if not yet connected, * initialize window to default value, and * initialize maxseg using peer's cached MSS. */ #ifdef INET6 if (isipv6) error = tcp6_connect(tp, sin6, td); #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET error = tcp_connect(tp, sinp, td); #endif /* * The bind operation in tcp_connect succeeded. We * no longer want to restore the flags if later * operations fail. */ if (error == 0 || inp->inp_lport != 0) restoreflags = false; if (error) { /* m is freed if PRUS_NOTREADY is unset. */ sbflush(&so->so_snd); goto out; } if (IS_FASTOPEN(tp->t_flags)) tcp_fastopen_connect(tp); else { tp->snd_wnd = TTCP_CLIENT_SND_WND; tcp_mss(tp, -1); } } if (flags & PRUS_EOF) { /* * Close the send side of the connection after * the data is sent. */ socantsendmore(so); tcp_usrclosed(tp); } if (TCPS_HAVEESTABLISHED(tp->t_state) && ((tp->t_flags2 & TF2_FBYTES_COMPLETE) == 0) && (tp->t_fbyte_out == 0) && (so->so_snd.sb_ccc > 0)) { tp->t_fbyte_out = ticks; if (tp->t_fbyte_out == 0) tp->t_fbyte_out = 1; if (tp->t_fbyte_out && tp->t_fbyte_in) tp->t_flags2 |= TF2_FBYTES_COMPLETE; } if (!(inp->inp_flags & INP_DROPPED) && !(flags & PRUS_NOTREADY)) { if (flags & PRUS_MORETOCOME) tp->t_flags |= TF_MORETOCOME; error = tcp_output_nodrop(tp); if (flags & PRUS_MORETOCOME) tp->t_flags &= ~TF_MORETOCOME; } } else { /* * XXXRW: PRUS_EOF not implemented with PRUS_OOB? */ SOCKBUF_LOCK(&so->so_snd); if (sbspace(&so->so_snd) < -512) { SOCKBUF_UNLOCK(&so->so_snd); error = ENOBUFS; goto out; } /* * According to RFC961 (Assigned Protocols), * the urgent pointer points to the last octet * of urgent data. We continue, however, * to consider it to indicate the first octet * of data past the urgent section. * Otherwise, snd_up should be one lower. */ if (tp->t_acktime == 0) tp->t_acktime = ticks; sbappendstream_locked(&so->so_snd, m, flags); SOCKBUF_UNLOCK(&so->so_snd); m = NULL; if (nam && tp->t_state < TCPS_SYN_SENT) { /* * Do implied connect if not yet connected, * initialize window to default value, and * initialize maxseg using peer's cached MSS. */ /* * Not going to contemplate SYN|URG */ if (IS_FASTOPEN(tp->t_flags)) tp->t_flags &= ~TF_FASTOPEN; #ifdef INET6 if (isipv6) error = tcp6_connect(tp, sin6, td); #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET error = tcp_connect(tp, sinp, td); #endif /* * The bind operation in tcp_connect succeeded. We * no longer want to restore the flags if later * operations fail. */ if (error == 0 || inp->inp_lport != 0) restoreflags = false; if (error != 0) { /* m is freed if PRUS_NOTREADY is unset. */ sbflush(&so->so_snd); goto out; } tp->snd_wnd = TTCP_CLIENT_SND_WND; tcp_mss(tp, -1); } tp->snd_up = tp->snd_una + sbavail(&so->so_snd); if ((flags & PRUS_NOTREADY) == 0) { tp->t_flags |= TF_FORCEDATA; error = tcp_output_nodrop(tp); tp->t_flags &= ~TF_FORCEDATA; } } TCP_LOG_EVENT(tp, NULL, &inp->inp_socket->so_rcv, &inp->inp_socket->so_snd, TCP_LOG_USERSEND, error, 0, NULL, false); out: /* * In case of PRUS_NOTREADY, the caller or tcp_usr_ready() is * responsible for freeing memory. */ if (m != NULL && (flags & PRUS_NOTREADY) == 0) m_freem(m); /* * If the request was unsuccessful and we changed flags, * restore the original flags. */ if (error != 0 && restoreflags) { inp->inp_vflag = vflagsav; inp->inp_inc.inc_flags = incflagsav; } tcp_bblog_pru(tp, (flags & PRUS_OOB) ? PRU_SENDOOB : ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND), error); TCP_PROBE2(debug__user, tp, (flags & PRUS_OOB) ? PRU_SENDOOB : ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND)); error = tcp_unlock_or_drop(tp, error); NET_EPOCH_EXIT(et); return (error); } static int tcp_usr_ready(struct socket *so, struct mbuf *m, int count) { struct epoch_tracker et; struct inpcb *inp; struct tcpcb *tp; int error; inp = sotoinpcb(so); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); mb_free_notready(m, count); return (ECONNRESET); } tp = intotcpcb(inp); SOCKBUF_LOCK(&so->so_snd); error = sbready(&so->so_snd, m, count); SOCKBUF_UNLOCK(&so->so_snd); if (error) { INP_WUNLOCK(inp); return (error); } NET_EPOCH_ENTER(et); error = tcp_output_unlock(tp); NET_EPOCH_EXIT(et); return (error); } /* * Abort the TCP. Drop the connection abruptly. */ static void tcp_usr_abort(struct socket *so) { struct inpcb *inp; struct tcpcb *tp; struct epoch_tracker et; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_abort: inp == NULL")); NET_EPOCH_ENTER(et); INP_WLOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_abort: inp_socket == NULL")); /* * If we still have full TCP state, and we're not dropped, drop. */ if (!(inp->inp_flags & INP_DROPPED)) { tp = intotcpcb(inp); tp = tcp_drop(tp, ECONNABORTED); if (tp == NULL) goto dropped; tcp_bblog_pru(tp, PRU_ABORT, 0); TCP_PROBE2(debug__user, tp, PRU_ABORT); } if (!(inp->inp_flags & INP_DROPPED)) { soref(so); inp->inp_flags |= INP_SOCKREF; } INP_WUNLOCK(inp); dropped: NET_EPOCH_EXIT(et); } /* * TCP socket is closed. Start friendly disconnect. */ static void tcp_usr_close(struct socket *so) { struct inpcb *inp; struct tcpcb *tp; struct epoch_tracker et; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_close: inp == NULL")); NET_EPOCH_ENTER(et); INP_WLOCK(inp); KASSERT(inp->inp_socket != NULL, ("tcp_usr_close: inp_socket == NULL")); /* * If we are still connected and we're not dropped, initiate * a disconnect. */ if (!(inp->inp_flags & INP_DROPPED)) { tp = intotcpcb(inp); if (tp->t_state != TCPS_TIME_WAIT) { tp->t_flags |= TF_CLOSED; tcp_disconnect(tp); tcp_bblog_pru(tp, PRU_CLOSE, 0); TCP_PROBE2(debug__user, tp, PRU_CLOSE); } } if (!(inp->inp_flags & INP_DROPPED)) { soref(so); inp->inp_flags |= INP_SOCKREF; } INP_WUNLOCK(inp); NET_EPOCH_EXIT(et); } static int tcp_pru_options_support(struct tcpcb *tp, int flags) { /* * If the specific TCP stack has a pru_options * specified then it does not always support * all the PRU_XX options and we must ask it. * If the function is not specified then all * of the PRU_XX options are supported. */ int ret = 0; if (tp->t_fb->tfb_pru_options) { ret = (*tp->t_fb->tfb_pru_options)(tp, flags); } return (ret); } /* * Receive out-of-band data. */ static int tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags) { int error = 0; struct inpcb *inp; struct tcpcb *tp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_usr_rcvoob: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (ECONNRESET); } tp = intotcpcb(inp); error = tcp_pru_options_support(tp, PRUS_OOB); if (error) { goto out; } if ((so->so_oobmark == 0 && (so->so_rcv.sb_state & SBS_RCVATMARK) == 0) || so->so_options & SO_OOBINLINE || tp->t_oobflags & TCPOOB_HADDATA) { error = EINVAL; goto out; } if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) { error = EWOULDBLOCK; goto out; } m->m_len = 1; *mtod(m, caddr_t) = tp->t_iobc; if ((flags & MSG_PEEK) == 0) tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA); out: tcp_bblog_pru(tp, PRU_RCVOOB, error); TCP_PROBE2(debug__user, tp, PRU_RCVOOB); INP_WUNLOCK(inp); return (error); } #ifdef INET struct protosw tcp_protosw = { .pr_type = SOCK_STREAM, .pr_protocol = IPPROTO_TCP, .pr_flags = PR_CONNREQUIRED | PR_IMPLOPCL | PR_WANTRCVD | PR_CAPATTACH, .pr_ctloutput = tcp_ctloutput, .pr_abort = tcp_usr_abort, .pr_accept = tcp_usr_accept, .pr_attach = tcp_usr_attach, .pr_bind = tcp_usr_bind, .pr_connect = tcp_usr_connect, .pr_control = in_control, .pr_detach = tcp_usr_detach, .pr_disconnect = tcp_usr_disconnect, .pr_listen = tcp_usr_listen, .pr_peeraddr = in_getpeeraddr, .pr_rcvd = tcp_usr_rcvd, .pr_rcvoob = tcp_usr_rcvoob, .pr_send = tcp_usr_send, .pr_ready = tcp_usr_ready, .pr_shutdown = tcp_usr_shutdown, .pr_sockaddr = in_getsockaddr, .pr_sosetlabel = in_pcbsosetlabel, .pr_close = tcp_usr_close, }; #endif /* INET */ #ifdef INET6 struct protosw tcp6_protosw = { .pr_type = SOCK_STREAM, .pr_protocol = IPPROTO_TCP, .pr_flags = PR_CONNREQUIRED | PR_IMPLOPCL |PR_WANTRCVD | PR_CAPATTACH, .pr_ctloutput = tcp_ctloutput, .pr_abort = tcp_usr_abort, .pr_accept = tcp6_usr_accept, .pr_attach = tcp_usr_attach, .pr_bind = tcp6_usr_bind, .pr_connect = tcp6_usr_connect, .pr_control = in6_control, .pr_detach = tcp_usr_detach, .pr_disconnect = tcp_usr_disconnect, .pr_listen = tcp6_usr_listen, .pr_peeraddr = in6_mapped_peeraddr, .pr_rcvd = tcp_usr_rcvd, .pr_rcvoob = tcp_usr_rcvoob, .pr_send = tcp_usr_send, .pr_ready = tcp_usr_ready, .pr_shutdown = tcp_usr_shutdown, .pr_sockaddr = in6_mapped_sockaddr, .pr_sosetlabel = in_pcbsosetlabel, .pr_close = tcp_usr_close, }; #endif /* INET6 */ #ifdef INET /* * Common subroutine to open a TCP connection to remote host specified * by struct sockaddr_in. Call in_pcbconnect() to choose local host address * and assign a local port number and install the inpcb into the hash. * Initialize connection parameters and enter SYN-SENT state. */ static int tcp_connect(struct tcpcb *tp, struct sockaddr_in *sin, struct thread *td) { struct inpcb *inp = tptoinpcb(tp); struct socket *so = tptosocket(tp); int error; NET_EPOCH_ASSERT(); INP_WLOCK_ASSERT(inp); if (__predict_false((so->so_state & (SS_ISCONNECTING | SS_ISCONNECTED | SS_ISDISCONNECTING | SS_ISDISCONNECTED)) != 0)) return (EISCONN); INP_HASH_WLOCK(&V_tcbinfo); error = in_pcbconnect(inp, sin, td->td_ucred, true); INP_HASH_WUNLOCK(&V_tcbinfo); if (error != 0) return (error); /* * Compute window scaling to request: * Scale to fit into sweet spot. See tcp_syncache.c. * XXX: This should move to tcp_output(). */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && (TCP_MAXWIN << tp->request_r_scale) < sb_max) tp->request_r_scale++; soisconnecting(so); TCPSTAT_INC(tcps_connattempt); tcp_state_change(tp, TCPS_SYN_SENT); tp->iss = tcp_new_isn(&inp->inp_inc); if (tp->t_flags & TF_REQ_TSTMP) tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); tcp_sendseqinit(tp); return (0); } #endif /* INET */ #ifdef INET6 static int tcp6_connect(struct tcpcb *tp, struct sockaddr_in6 *sin6, struct thread *td) { struct inpcb *inp = tptoinpcb(tp); struct socket *so = tptosocket(tp); int error; NET_EPOCH_ASSERT(); INP_WLOCK_ASSERT(inp); if (__predict_false((so->so_state & (SS_ISCONNECTING | SS_ISCONNECTED)) != 0)) return (EISCONN); INP_HASH_WLOCK(&V_tcbinfo); error = in6_pcbconnect(inp, sin6, td->td_ucred, true); INP_HASH_WUNLOCK(&V_tcbinfo); if (error != 0) return (error); /* Compute window scaling to request. */ while (tp->request_r_scale < TCP_MAX_WINSHIFT && (TCP_MAXWIN << tp->request_r_scale) < sb_max) tp->request_r_scale++; soisconnecting(so); TCPSTAT_INC(tcps_connattempt); tcp_state_change(tp, TCPS_SYN_SENT); tp->iss = tcp_new_isn(&inp->inp_inc); if (tp->t_flags & TF_REQ_TSTMP) tp->ts_offset = tcp_new_ts_offset(&inp->inp_inc); tcp_sendseqinit(tp); return (0); } #endif /* INET6 */ /* * Export TCP internal state information via a struct tcp_info, based on the * Linux 2.6 API. Not ABI compatible as our constants are mapped differently * (TCP state machine, etc). We export all information using FreeBSD-native * constants -- for example, the numeric values for tcpi_state will differ * from Linux. */ void tcp_fill_info(const struct tcpcb *tp, struct tcp_info *ti) { INP_LOCK_ASSERT(tptoinpcb(tp)); bzero(ti, sizeof(*ti)); ti->tcpi_state = tp->t_state; if ((tp->t_flags & TF_REQ_TSTMP) && (tp->t_flags & TF_RCVD_TSTMP)) ti->tcpi_options |= TCPI_OPT_TIMESTAMPS; if (tp->t_flags & TF_SACK_PERMIT) ti->tcpi_options |= TCPI_OPT_SACK; if ((tp->t_flags & TF_REQ_SCALE) && (tp->t_flags & TF_RCVD_SCALE)) { ti->tcpi_options |= TCPI_OPT_WSCALE; ti->tcpi_snd_wscale = tp->snd_scale; ti->tcpi_rcv_wscale = tp->rcv_scale; } switch (tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) { case TF2_ECN_PERMIT: ti->tcpi_options |= TCPI_OPT_ECN; break; case TF2_ACE_PERMIT: /* FALLTHROUGH */ case TF2_ECN_PERMIT | TF2_ACE_PERMIT: ti->tcpi_options |= TCPI_OPT_ACE; break; default: break; } if (IS_FASTOPEN(tp->t_flags)) ti->tcpi_options |= TCPI_OPT_TFO; ti->tcpi_rto = tp->t_rxtcur * tick; ti->tcpi_last_data_recv = ((uint32_t)ticks - tp->t_rcvtime) * tick; ti->tcpi_rtt = ((u_int64_t)tp->t_srtt * tick) >> TCP_RTT_SHIFT; ti->tcpi_rttvar = ((u_int64_t)tp->t_rttvar * tick) >> TCP_RTTVAR_SHIFT; ti->tcpi_snd_ssthresh = tp->snd_ssthresh; ti->tcpi_snd_cwnd = tp->snd_cwnd; /* * FreeBSD-specific extension fields for tcp_info. */ ti->tcpi_rcv_space = tp->rcv_wnd; ti->tcpi_rcv_nxt = tp->rcv_nxt; ti->tcpi_snd_wnd = tp->snd_wnd; ti->tcpi_snd_bwnd = 0; /* Unused, kept for compat. */ ti->tcpi_snd_nxt = tp->snd_nxt; ti->tcpi_snd_mss = tp->t_maxseg; ti->tcpi_rcv_mss = tp->t_maxseg; ti->tcpi_snd_rexmitpack = tp->t_sndrexmitpack; ti->tcpi_rcv_ooopack = tp->t_rcvoopack; ti->tcpi_snd_zerowin = tp->t_sndzerowin; ti->tcpi_snd_una = tp->snd_una; ti->tcpi_snd_max = tp->snd_max; ti->tcpi_rcv_numsacks = tp->rcv_numsacks; ti->tcpi_rcv_adv = tp->rcv_adv; ti->tcpi_dupacks = tp->t_dupacks; #ifdef TCP_OFFLOAD if (tp->t_flags & TF_TOE) { ti->tcpi_options |= TCPI_OPT_TOE; tcp_offload_tcp_info(tp, ti); } #endif /* * AccECN related counters. */ if ((tp->t_flags2 & (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) == (TF2_ECN_PERMIT | TF2_ACE_PERMIT)) /* * Internal counter starts at 5 for AccECN * but 0 for RFC3168 ECN. */ ti->tcpi_delivered_ce = tp->t_scep - 5; else ti->tcpi_delivered_ce = tp->t_scep; ti->tcpi_received_ce = tp->t_rcep; } /* * tcp_ctloutput() must drop the inpcb lock before performing copyin on * socket option arguments. When it re-acquires the lock after the copy, it * has to revalidate that the connection is still valid for the socket * option. */ #define INP_WLOCK_RECHECK_CLEANUP(inp, cleanup) do { \ INP_WLOCK(inp); \ if (inp->inp_flags & INP_DROPPED) { \ INP_WUNLOCK(inp); \ cleanup; \ return (ECONNRESET); \ } \ tp = intotcpcb(inp); \ } while(0) #define INP_WLOCK_RECHECK(inp) INP_WLOCK_RECHECK_CLEANUP((inp), /* noop */) int tcp_ctloutput_set(struct inpcb *inp, struct sockopt *sopt) { struct socket *so = inp->inp_socket; struct tcpcb *tp = intotcpcb(inp); int error = 0; MPASS(sopt->sopt_dir == SOPT_SET); INP_WLOCK_ASSERT(inp); KASSERT((inp->inp_flags & INP_DROPPED) == 0, ("inp_flags == %x", inp->inp_flags)); KASSERT(so != NULL, ("inp_socket == NULL")); if (sopt->sopt_level != IPPROTO_TCP) { INP_WUNLOCK(inp); #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) error = ip6_ctloutput(so, sopt); #endif #if defined(INET6) && defined(INET) else #endif #ifdef INET error = ip_ctloutput(so, sopt); #endif /* * When an IP-level socket option affects TCP, pass control * down to stack tfb_tcp_ctloutput, otherwise return what * IP level returned. */ switch (sopt->sopt_level) { #ifdef INET6 case IPPROTO_IPV6: if ((inp->inp_vflag & INP_IPV6PROTO) == 0) return (error); switch (sopt->sopt_name) { case IPV6_TCLASS: /* Notify tcp stacks that care (e.g. RACK). */ break; case IPV6_USE_MIN_MTU: /* Update t_maxseg accordingly. */ break; default: return (error); } break; #endif #ifdef INET case IPPROTO_IP: switch (sopt->sopt_name) { case IP_TOS: inp->inp_ip_tos &= ~IPTOS_ECN_MASK; break; case IP_TTL: /* Notify tcp stacks that care (e.g. RACK). */ break; default: return (error); } break; #endif default: return (error); } INP_WLOCK_RECHECK(inp); } else if (sopt->sopt_name == TCP_FUNCTION_BLK) { /* * Protect the TCP option TCP_FUNCTION_BLK so * that a sub-function can *never* overwrite this. */ struct tcp_function_set fsn; struct tcp_function_block *blk; void *ptr = NULL; INP_WUNLOCK(inp); error = sooptcopyin(sopt, &fsn, sizeof fsn, sizeof fsn); if (error) return (error); INP_WLOCK_RECHECK(inp); blk = find_and_ref_tcp_functions(&fsn); if (blk == NULL) { INP_WUNLOCK(inp); return (ENOENT); } if (tp->t_fb == blk) { /* You already have this */ refcount_release(&blk->tfb_refcnt); INP_WUNLOCK(inp); return (0); } if (blk->tfb_flags & TCP_FUNC_BEING_REMOVED) { refcount_release(&blk->tfb_refcnt); INP_WUNLOCK(inp); return (ENOENT); } error = (*blk->tfb_tcp_handoff_ok)(tp); if (error) { refcount_release(&blk->tfb_refcnt); INP_WUNLOCK(inp); return (error); } /* * Ensure the new stack takes ownership with a * clean slate on peak rate threshold. */ if (tp->t_fb->tfb_tcp_timer_stop_all != NULL) tp->t_fb->tfb_tcp_timer_stop_all(tp); if (blk->tfb_tcp_fb_init) { error = (*blk->tfb_tcp_fb_init)(tp, &ptr); if (error) { /* * Release the ref count the lookup * acquired. */ refcount_release(&blk->tfb_refcnt); /* * Now there is a chance that the * init() function mucked with some * things before it failed, such as * hpts or inp_flags2 or timer granularity. * It should not of, but lets give the old * stack a chance to reset to a known good state. */ if (tp->t_fb->tfb_switch_failed) { (*tp->t_fb->tfb_switch_failed)(tp); } goto err_out; } } if (tp->t_fb->tfb_tcp_fb_fini) { struct epoch_tracker et; /* * Tell the stack to cleanup with 0 i.e. * the tcb is not going away. */ NET_EPOCH_ENTER(et); (*tp->t_fb->tfb_tcp_fb_fini)(tp, 0); NET_EPOCH_EXIT(et); } /* * Release the old refcnt, the * lookup acquired a ref on the * new one already. */ refcount_release(&tp->t_fb->tfb_refcnt); /* * Set in the new stack. */ tp->t_fb = blk; tp->t_fb_ptr = ptr; #ifdef TCP_OFFLOAD if (tp->t_flags & TF_TOE) { tcp_offload_ctloutput(tp, sopt->sopt_dir, sopt->sopt_name); } #endif err_out: INP_WUNLOCK(inp); return (error); } /* Pass in the INP locked, callee must unlock it. */ return (tp->t_fb->tfb_tcp_ctloutput(tp, sopt)); } static int tcp_ctloutput_get(struct inpcb *inp, struct sockopt *sopt) { struct socket *so = inp->inp_socket; struct tcpcb *tp = intotcpcb(inp); int error = 0; MPASS(sopt->sopt_dir == SOPT_GET); INP_WLOCK_ASSERT(inp); KASSERT((inp->inp_flags & INP_DROPPED) == 0, ("inp_flags == %x", inp->inp_flags)); KASSERT(so != NULL, ("inp_socket == NULL")); if (sopt->sopt_level != IPPROTO_TCP) { INP_WUNLOCK(inp); #ifdef INET6 if (inp->inp_vflag & INP_IPV6PROTO) error = ip6_ctloutput(so, sopt); #endif /* INET6 */ #if defined(INET6) && defined(INET) else #endif #ifdef INET error = ip_ctloutput(so, sopt); #endif return (error); } if (((sopt->sopt_name == TCP_FUNCTION_BLK) || (sopt->sopt_name == TCP_FUNCTION_ALIAS))) { struct tcp_function_set fsn; if (sopt->sopt_name == TCP_FUNCTION_ALIAS) { memset(&fsn, 0, sizeof(fsn)); find_tcp_function_alias(tp->t_fb, &fsn); } else { strncpy(fsn.function_set_name, tp->t_fb->tfb_tcp_block_name, TCP_FUNCTION_NAME_LEN_MAX); fsn.function_set_name[TCP_FUNCTION_NAME_LEN_MAX - 1] = '\0'; } fsn.pcbcnt = tp->t_fb->tfb_refcnt; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &fsn, sizeof fsn); return (error); } /* Pass in the INP locked, callee must unlock it. */ return (tp->t_fb->tfb_tcp_ctloutput(tp, sopt)); } int tcp_ctloutput(struct socket *so, struct sockopt *sopt) { struct inpcb *inp; inp = sotoinpcb(so); KASSERT(inp != NULL, ("tcp_ctloutput: inp == NULL")); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); return (ECONNRESET); } if (sopt->sopt_dir == SOPT_SET) return (tcp_ctloutput_set(inp, sopt)); else if (sopt->sopt_dir == SOPT_GET) return (tcp_ctloutput_get(inp, sopt)); else panic("%s: sopt_dir $%d", __func__, sopt->sopt_dir); } /* * If this assert becomes untrue, we need to change the size of the buf * variable in tcp_default_ctloutput(). */ #ifdef CTASSERT CTASSERT(TCP_CA_NAME_MAX <= TCP_LOG_ID_LEN); CTASSERT(TCP_LOG_REASON_LEN <= TCP_LOG_ID_LEN); #endif #ifdef KERN_TLS static int copyin_tls_enable(struct sockopt *sopt, struct tls_enable *tls) { struct tls_enable_v0 tls_v0; int error; if (sopt->sopt_valsize == sizeof(tls_v0)) { error = sooptcopyin(sopt, &tls_v0, sizeof(tls_v0), sizeof(tls_v0)); if (error) return (error); memset(tls, 0, sizeof(*tls)); tls->cipher_key = tls_v0.cipher_key; tls->iv = tls_v0.iv; tls->auth_key = tls_v0.auth_key; tls->cipher_algorithm = tls_v0.cipher_algorithm; tls->cipher_key_len = tls_v0.cipher_key_len; tls->iv_len = tls_v0.iv_len; tls->auth_algorithm = tls_v0.auth_algorithm; tls->auth_key_len = tls_v0.auth_key_len; tls->flags = tls_v0.flags; tls->tls_vmajor = tls_v0.tls_vmajor; tls->tls_vminor = tls_v0.tls_vminor; return (0); } return (sooptcopyin(sopt, tls, sizeof(*tls), sizeof(*tls))); } #endif extern struct cc_algo newreno_cc_algo; static int tcp_set_cc_mod(struct inpcb *inp, struct sockopt *sopt) { struct cc_algo *algo; void *ptr = NULL; struct tcpcb *tp; struct cc_var cc_mem; char buf[TCP_CA_NAME_MAX]; size_t mem_sz; int error; INP_WUNLOCK(inp); error = sooptcopyin(sopt, buf, TCP_CA_NAME_MAX - 1, 1); if (error) return(error); buf[sopt->sopt_valsize] = '\0'; CC_LIST_RLOCK(); STAILQ_FOREACH(algo, &cc_list, entries) { if (strncmp(buf, algo->name, TCP_CA_NAME_MAX) == 0) { if (algo->flags & CC_MODULE_BEING_REMOVED) { /* We can't "see" modules being unloaded */ continue; } break; } } if (algo == NULL) { CC_LIST_RUNLOCK(); return(ESRCH); } /* * With a reference the algorithm cannot be removed * so we hold a reference through the change process. */ cc_refer(algo); CC_LIST_RUNLOCK(); if (algo->cb_init != NULL) { /* We can now pre-get the memory for the CC */ mem_sz = (*algo->cc_data_sz)(); if (mem_sz == 0) { goto no_mem_needed; } ptr = malloc(mem_sz, M_CC_MEM, M_WAITOK); } else { no_mem_needed: mem_sz = 0; ptr = NULL; } /* * Make sure its all clean and zero and also get * back the inplock. */ memset(&cc_mem, 0, sizeof(cc_mem)); INP_WLOCK(inp); if (inp->inp_flags & INP_DROPPED) { INP_WUNLOCK(inp); if (ptr) free(ptr, M_CC_MEM); /* Release our temp reference */ CC_LIST_RLOCK(); cc_release(algo); CC_LIST_RUNLOCK(); return (ECONNRESET); } tp = intotcpcb(inp); if (ptr != NULL) memset(ptr, 0, mem_sz); cc_mem.ccvc.tcp = tp; /* * We once again hold a write lock over the tcb so it's * safe to do these things without ordering concerns. * Note here we init into stack memory. */ if (algo->cb_init != NULL) error = algo->cb_init(&cc_mem, ptr); else error = 0; /* * The CC algorithms, when given their memory * should not fail we could in theory have a * KASSERT here. */ if (error == 0) { /* * Touchdown, lets go ahead and move the * connection to the new CC module by * copying in the cc_mem after we call * the old ones cleanup (if any). */ if (CC_ALGO(tp)->cb_destroy != NULL) CC_ALGO(tp)->cb_destroy(&tp->t_ccv); /* Detach the old CC from the tcpcb */ cc_detach(tp); /* Copy in our temp memory that was inited */ memcpy(&tp->t_ccv, &cc_mem, sizeof(struct cc_var)); /* Now attach the new, which takes a reference */ cc_attach(tp, algo); /* Ok now are we where we have gotten past any conn_init? */ if (TCPS_HAVEESTABLISHED(tp->t_state) && (CC_ALGO(tp)->conn_init != NULL)) { /* Yep run the connection init for the new CC */ CC_ALGO(tp)->conn_init(&tp->t_ccv); } } else if (ptr) free(ptr, M_CC_MEM); INP_WUNLOCK(inp); /* Now lets release our temp reference */ CC_LIST_RLOCK(); cc_release(algo); CC_LIST_RUNLOCK(); return (error); } int tcp_default_ctloutput(struct tcpcb *tp, struct sockopt *sopt) { struct inpcb *inp = tptoinpcb(tp); int error, opt, optval; u_int ui; struct tcp_info ti; #ifdef KERN_TLS struct tls_enable tls; struct socket *so = inp->inp_socket; #endif char *pbuf, buf[TCP_LOG_ID_LEN]; #ifdef STATS struct statsblob *sbp; #endif size_t len; INP_WLOCK_ASSERT(inp); KASSERT((inp->inp_flags & INP_DROPPED) == 0, ("inp_flags == %x", inp->inp_flags)); KASSERT(inp->inp_socket != NULL, ("inp_socket == NULL")); switch (sopt->sopt_level) { #ifdef INET6 case IPPROTO_IPV6: MPASS(inp->inp_vflag & INP_IPV6PROTO); switch (sopt->sopt_name) { case IPV6_USE_MIN_MTU: tcp6_use_min_mtu(tp); /* FALLTHROUGH */ } INP_WUNLOCK(inp); return (0); #endif #ifdef INET case IPPROTO_IP: INP_WUNLOCK(inp); return (0); #endif } /* * For TCP_CCALGOOPT forward the control to CC module, for both * SOPT_SET and SOPT_GET. */ switch (sopt->sopt_name) { case TCP_CCALGOOPT: INP_WUNLOCK(inp); if (sopt->sopt_valsize > CC_ALGOOPT_LIMIT) return (EINVAL); pbuf = malloc(sopt->sopt_valsize, M_TEMP, M_WAITOK | M_ZERO); error = sooptcopyin(sopt, pbuf, sopt->sopt_valsize, sopt->sopt_valsize); if (error) { free(pbuf, M_TEMP); return (error); } INP_WLOCK_RECHECK_CLEANUP(inp, free(pbuf, M_TEMP)); if (CC_ALGO(tp)->ctl_output != NULL) error = CC_ALGO(tp)->ctl_output(&tp->t_ccv, sopt, pbuf); else error = ENOENT; INP_WUNLOCK(inp); if (error == 0 && sopt->sopt_dir == SOPT_GET) error = sooptcopyout(sopt, pbuf, sopt->sopt_valsize); free(pbuf, M_TEMP); return (error); } switch (sopt->sopt_dir) { case SOPT_SET: switch (sopt->sopt_name) { #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE) case TCP_MD5SIG: INP_WUNLOCK(inp); if (!TCPMD5_ENABLED()) return (ENOPROTOOPT); error = TCPMD5_PCBCTL(inp, sopt); if (error) return (error); INP_WLOCK_RECHECK(inp); goto unlock_and_done; #endif /* IPSEC */ case TCP_NODELAY: case TCP_NOOPT: case TCP_LRD: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); INP_WLOCK_RECHECK(inp); switch (sopt->sopt_name) { case TCP_NODELAY: opt = TF_NODELAY; break; case TCP_NOOPT: opt = TF_NOOPT; break; case TCP_LRD: opt = TF_LRD; break; default: opt = 0; /* dead code to fool gcc */ break; } if (optval) tp->t_flags |= opt; else tp->t_flags &= ~opt; unlock_and_done: #ifdef TCP_OFFLOAD if (tp->t_flags & TF_TOE) { tcp_offload_ctloutput(tp, sopt->sopt_dir, sopt->sopt_name); } #endif INP_WUNLOCK(inp); break; case TCP_NOPUSH: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); INP_WLOCK_RECHECK(inp); if (optval) tp->t_flags |= TF_NOPUSH; else if (tp->t_flags & TF_NOPUSH) { tp->t_flags &= ~TF_NOPUSH; if (TCPS_HAVEESTABLISHED(tp->t_state)) { struct epoch_tracker et; NET_EPOCH_ENTER(et); error = tcp_output_nodrop(tp); NET_EPOCH_EXIT(et); } } goto unlock_and_done; case TCP_REMOTE_UDP_ENCAPS_PORT: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); if ((optval < TCP_TUNNELING_PORT_MIN) || (optval > TCP_TUNNELING_PORT_MAX)) { /* Its got to be in range */ return (EINVAL); } if ((V_tcp_udp_tunneling_port == 0) && (optval != 0)) { /* You have to have enabled a UDP tunneling port first */ return (EINVAL); } INP_WLOCK_RECHECK(inp); if (tp->t_state != TCPS_CLOSED) { /* You can't change after you are connected */ error = EINVAL; } else { /* Ok we are all good set the port */ tp->t_port = htons(optval); } goto unlock_and_done; case TCP_MAXSEG: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); INP_WLOCK_RECHECK(inp); if (optval > 0 && optval <= tp->t_maxseg && optval + 40 >= V_tcp_minmss) tp->t_maxseg = optval; else error = EINVAL; goto unlock_and_done; case TCP_INFO: INP_WUNLOCK(inp); error = EINVAL; break; case TCP_STATS: INP_WUNLOCK(inp); #ifdef STATS error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); if (optval > 0) sbp = stats_blob_alloc( V_tcp_perconn_stats_dflt_tpl, 0); else sbp = NULL; INP_WLOCK_RECHECK(inp); if ((tp->t_stats != NULL && sbp == NULL) || (tp->t_stats == NULL && sbp != NULL)) { struct statsblob *t = tp->t_stats; tp->t_stats = sbp; sbp = t; } INP_WUNLOCK(inp); stats_blob_destroy(sbp); #else return (EOPNOTSUPP); #endif /* !STATS */ break; case TCP_CONGESTION: error = tcp_set_cc_mod(inp, sopt); break; case TCP_REUSPORT_LB_NUMA: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof(optval), sizeof(optval)); INP_WLOCK_RECHECK(inp); if (!error) error = in_pcblbgroup_numa(inp, optval); INP_WUNLOCK(inp); break; #ifdef KERN_TLS case TCP_TXTLS_ENABLE: INP_WUNLOCK(inp); error = copyin_tls_enable(sopt, &tls); if (error) break; error = ktls_enable_tx(so, &tls); break; case TCP_TXTLS_MODE: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); if (error) return (error); INP_WLOCK_RECHECK(inp); error = ktls_set_tx_mode(so, ui); INP_WUNLOCK(inp); break; case TCP_RXTLS_ENABLE: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &tls, sizeof(tls), sizeof(tls)); if (error) break; error = ktls_enable_rx(so, &tls); break; #endif case TCP_MAXUNACKTIME: case TCP_KEEPIDLE: case TCP_KEEPINTVL: case TCP_KEEPINIT: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); if (error) return (error); if (ui > (UINT_MAX / hz)) { error = EINVAL; break; } ui *= hz; INP_WLOCK_RECHECK(inp); switch (sopt->sopt_name) { case TCP_MAXUNACKTIME: tp->t_maxunacktime = ui; break; case TCP_KEEPIDLE: tp->t_keepidle = ui; /* * XXX: better check current remaining * timeout and "merge" it with new value. */ if ((tp->t_state > TCPS_LISTEN) && (tp->t_state <= TCPS_CLOSING)) tcp_timer_activate(tp, TT_KEEP, TP_KEEPIDLE(tp)); break; case TCP_KEEPINTVL: tp->t_keepintvl = ui; if ((tp->t_state == TCPS_FIN_WAIT_2) && (TP_MAXIDLE(tp) > 0)) tcp_timer_activate(tp, TT_2MSL, TP_MAXIDLE(tp)); break; case TCP_KEEPINIT: tp->t_keepinit = ui; if (tp->t_state == TCPS_SYN_RECEIVED || tp->t_state == TCPS_SYN_SENT) tcp_timer_activate(tp, TT_KEEP, TP_KEEPINIT(tp)); break; } goto unlock_and_done; case TCP_KEEPCNT: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &ui, sizeof(ui), sizeof(ui)); if (error) return (error); INP_WLOCK_RECHECK(inp); tp->t_keepcnt = ui; if ((tp->t_state == TCPS_FIN_WAIT_2) && (TP_MAXIDLE(tp) > 0)) tcp_timer_activate(tp, TT_2MSL, TP_MAXIDLE(tp)); goto unlock_and_done; #ifdef TCPPCAP case TCP_PCAP_OUT: case TCP_PCAP_IN: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); INP_WLOCK_RECHECK(inp); if (optval >= 0) tcp_pcap_set_sock_max( (sopt->sopt_name == TCP_PCAP_OUT) ? &(tp->t_outpkts) : &(tp->t_inpkts), optval); else error = EINVAL; goto unlock_and_done; #endif case TCP_FASTOPEN: { struct tcp_fastopen tfo_optval; INP_WUNLOCK(inp); if (!V_tcp_fastopen_client_enable && !V_tcp_fastopen_server_enable) return (EPERM); error = sooptcopyin(sopt, &tfo_optval, sizeof(tfo_optval), sizeof(int)); if (error) return (error); INP_WLOCK_RECHECK(inp); if ((tp->t_state != TCPS_CLOSED) && (tp->t_state != TCPS_LISTEN)) { error = EINVAL; goto unlock_and_done; } if (tfo_optval.enable) { if (tp->t_state == TCPS_LISTEN) { if (!V_tcp_fastopen_server_enable) { error = EPERM; goto unlock_and_done; } if (tp->t_tfo_pending == NULL) tp->t_tfo_pending = tcp_fastopen_alloc_counter(); } else { /* * If a pre-shared key was provided, * stash it in the client cookie * field of the tcpcb for use during * connect. */ if (sopt->sopt_valsize == sizeof(tfo_optval)) { memcpy(tp->t_tfo_cookie.client, tfo_optval.psk, TCP_FASTOPEN_PSK_LEN); tp->t_tfo_client_cookie_len = TCP_FASTOPEN_PSK_LEN; } } tp->t_flags |= TF_FASTOPEN; } else tp->t_flags &= ~TF_FASTOPEN; goto unlock_and_done; } #ifdef TCP_BLACKBOX case TCP_LOG: INP_WUNLOCK(inp); error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); if (error) return (error); INP_WLOCK_RECHECK(inp); error = tcp_log_state_change(tp, optval); goto unlock_and_done; case TCP_LOGBUF: INP_WUNLOCK(inp); error = EINVAL; break; case TCP_LOGID: INP_WUNLOCK(inp); error = sooptcopyin(sopt, buf, TCP_LOG_ID_LEN - 1, 0); if (error) break; buf[sopt->sopt_valsize] = '\0'; INP_WLOCK_RECHECK(inp); error = tcp_log_set_id(tp, buf); /* tcp_log_set_id() unlocks the INP. */ break; case TCP_LOGDUMP: case TCP_LOGDUMPID: INP_WUNLOCK(inp); error = sooptcopyin(sopt, buf, TCP_LOG_REASON_LEN - 1, 0); if (error) break; buf[sopt->sopt_valsize] = '\0'; INP_WLOCK_RECHECK(inp); if (sopt->sopt_name == TCP_LOGDUMP) { error = tcp_log_dump_tp_logbuf(tp, buf, M_WAITOK, true); INP_WUNLOCK(inp); } else { tcp_log_dump_tp_bucket_logbufs(tp, buf); /* * tcp_log_dump_tp_bucket_logbufs() drops the * INP lock. */ } break; #endif default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; case SOPT_GET: tp = intotcpcb(inp); switch (sopt->sopt_name) { #if defined(IPSEC_SUPPORT) || defined(TCP_SIGNATURE) case TCP_MD5SIG: INP_WUNLOCK(inp); if (!TCPMD5_ENABLED()) return (ENOPROTOOPT); error = TCPMD5_PCBCTL(inp, sopt); break; #endif case TCP_NODELAY: optval = tp->t_flags & TF_NODELAY; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_MAXSEG: optval = tp->t_maxseg; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_REMOTE_UDP_ENCAPS_PORT: optval = ntohs(tp->t_port); INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_NOOPT: optval = tp->t_flags & TF_NOOPT; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_NOPUSH: optval = tp->t_flags & TF_NOPUSH; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; case TCP_INFO: tcp_fill_info(tp, &ti); INP_WUNLOCK(inp); error = sooptcopyout(sopt, &ti, sizeof ti); break; case TCP_STATS: { #ifdef STATS int nheld; TYPEOF_MEMBER(struct statsblob, flags) sbflags = 0; error = 0; socklen_t outsbsz = sopt->sopt_valsize; if (tp->t_stats == NULL) error = ENOENT; else if (outsbsz >= tp->t_stats->cursz) outsbsz = tp->t_stats->cursz; else if (outsbsz >= sizeof(struct statsblob)) outsbsz = sizeof(struct statsblob); else error = EINVAL; INP_WUNLOCK(inp); if (error) break; sbp = sopt->sopt_val; nheld = atop(round_page(((vm_offset_t)sbp) + (vm_size_t)outsbsz) - trunc_page((vm_offset_t)sbp)); vm_page_t ma[nheld]; if (vm_fault_quick_hold_pages( &curproc->p_vmspace->vm_map, (vm_offset_t)sbp, outsbsz, VM_PROT_READ | VM_PROT_WRITE, ma, nheld) < 0) { error = EFAULT; break; } if ((error = copyin_nofault(&(sbp->flags), &sbflags, SIZEOF_MEMBER(struct statsblob, flags)))) goto unhold; INP_WLOCK_RECHECK(inp); error = stats_blob_snapshot(&sbp, outsbsz, tp->t_stats, sbflags | SB_CLONE_USRDSTNOFAULT); INP_WUNLOCK(inp); sopt->sopt_valsize = outsbsz; unhold: vm_page_unhold_pages(ma, nheld); #else INP_WUNLOCK(inp); error = EOPNOTSUPP; #endif /* !STATS */ break; } case TCP_CONGESTION: len = strlcpy(buf, CC_ALGO(tp)->name, TCP_CA_NAME_MAX); INP_WUNLOCK(inp); error = sooptcopyout(sopt, buf, len + 1); break; case TCP_MAXUNACKTIME: case TCP_KEEPIDLE: case TCP_KEEPINTVL: case TCP_KEEPINIT: case TCP_KEEPCNT: switch (sopt->sopt_name) { case TCP_MAXUNACKTIME: ui = TP_MAXUNACKTIME(tp) / hz; break; case TCP_KEEPIDLE: ui = TP_KEEPIDLE(tp) / hz; break; case TCP_KEEPINTVL: ui = TP_KEEPINTVL(tp) / hz; break; case TCP_KEEPINIT: ui = TP_KEEPINIT(tp) / hz; break; case TCP_KEEPCNT: ui = TP_KEEPCNT(tp); break; } INP_WUNLOCK(inp); error = sooptcopyout(sopt, &ui, sizeof(ui)); break; #ifdef TCPPCAP case TCP_PCAP_OUT: case TCP_PCAP_IN: optval = tcp_pcap_get_sock_max( (sopt->sopt_name == TCP_PCAP_OUT) ? &(tp->t_outpkts) : &(tp->t_inpkts)); INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; #endif case TCP_FASTOPEN: optval = tp->t_flags & TF_FASTOPEN; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; #ifdef TCP_BLACKBOX case TCP_LOG: optval = tcp_get_bblog_state(tp); INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof(optval)); break; case TCP_LOGBUF: /* tcp_log_getlogbuf() does INP_WUNLOCK(inp) */ error = tcp_log_getlogbuf(sopt, tp); break; case TCP_LOGID: len = tcp_log_get_id(tp, buf); INP_WUNLOCK(inp); error = sooptcopyout(sopt, buf, len + 1); break; case TCP_LOGDUMP: case TCP_LOGDUMPID: INP_WUNLOCK(inp); error = EINVAL; break; #endif #ifdef KERN_TLS case TCP_TXTLS_MODE: error = ktls_get_tx_mode(so, &optval); INP_WUNLOCK(inp); if (error == 0) error = sooptcopyout(sopt, &optval, sizeof(optval)); break; case TCP_RXTLS_MODE: error = ktls_get_rx_mode(so, &optval); INP_WUNLOCK(inp); if (error == 0) error = sooptcopyout(sopt, &optval, sizeof(optval)); break; #endif case TCP_LRD: optval = tp->t_flags & TF_LRD; INP_WUNLOCK(inp); error = sooptcopyout(sopt, &optval, sizeof optval); break; default: INP_WUNLOCK(inp); error = ENOPROTOOPT; break; } break; } return (error); } #undef INP_WLOCK_RECHECK #undef INP_WLOCK_RECHECK_CLEANUP /* * Initiate (or continue) disconnect. * If embryonic state, just send reset (once). * If in ``let data drain'' option and linger null, just drop. * Otherwise (hard), mark socket disconnecting and drop * current input data; switch states based on user close, and * send segment to peer (with FIN). */ static void tcp_disconnect(struct tcpcb *tp) { struct inpcb *inp = tptoinpcb(tp); struct socket *so = tptosocket(tp); NET_EPOCH_ASSERT(); INP_WLOCK_ASSERT(inp); /* * Neither tcp_close() nor tcp_drop() should return NULL, as the * socket is still open. */ if (tp->t_state < TCPS_ESTABLISHED && !(tp->t_state > TCPS_LISTEN && IS_FASTOPEN(tp->t_flags))) { tp = tcp_close(tp); KASSERT(tp != NULL, ("tcp_disconnect: tcp_close() returned NULL")); } else if ((so->so_options & SO_LINGER) && so->so_linger == 0) { tp = tcp_drop(tp, 0); KASSERT(tp != NULL, ("tcp_disconnect: tcp_drop() returned NULL")); } else { soisdisconnecting(so); sbflush(&so->so_rcv); tcp_usrclosed(tp); if (!(inp->inp_flags & INP_DROPPED)) /* Ignore stack's drop request, we already at it. */ (void)tcp_output_nodrop(tp); } } /* * User issued close, and wish to trail through shutdown states: * if never received SYN, just forget it. If got a SYN from peer, * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN. * If already got a FIN from peer, then almost done; go to LAST_ACK * state. In all other cases, have already sent FIN to peer (e.g. * after PRU_SHUTDOWN), and just have to play tedious game waiting * for peer to send FIN or not respond to keep-alives, etc. * We can let the user exit from the close as soon as the FIN is acked. */ static void tcp_usrclosed(struct tcpcb *tp) { NET_EPOCH_ASSERT(); INP_WLOCK_ASSERT(tptoinpcb(tp)); switch (tp->t_state) { case TCPS_LISTEN: #ifdef TCP_OFFLOAD tcp_offload_listen_stop(tp); #endif tcp_state_change(tp, TCPS_CLOSED); /* FALLTHROUGH */ case TCPS_CLOSED: tp = tcp_close(tp); /* * tcp_close() should never return NULL here as the socket is * still open. */ KASSERT(tp != NULL, ("tcp_usrclosed: tcp_close() returned NULL")); break; case TCPS_SYN_SENT: case TCPS_SYN_RECEIVED: tp->t_flags |= TF_NEEDFIN; break; case TCPS_ESTABLISHED: tcp_state_change(tp, TCPS_FIN_WAIT_1); break; case TCPS_CLOSE_WAIT: tcp_state_change(tp, TCPS_LAST_ACK); break; } if (tp->t_acktime == 0) tp->t_acktime = ticks; if (tp->t_state >= TCPS_FIN_WAIT_2) { soisdisconnected(tptosocket(tp)); /* Prevent the connection hanging in FIN_WAIT_2 forever. */ if (tp->t_state == TCPS_FIN_WAIT_2) { int timeout; timeout = (tcp_fast_finwait2_recycle) ? tcp_finwait2_timeout : TP_MAXIDLE(tp); tcp_timer_activate(tp, TT_2MSL, timeout); } } } #ifdef DDB static void db_print_indent(int indent) { int i; for (i = 0; i < indent; i++) db_printf(" "); } static void db_print_tstate(int t_state) { switch (t_state) { case TCPS_CLOSED: db_printf("TCPS_CLOSED"); return; case TCPS_LISTEN: db_printf("TCPS_LISTEN"); return; case TCPS_SYN_SENT: db_printf("TCPS_SYN_SENT"); return; case TCPS_SYN_RECEIVED: db_printf("TCPS_SYN_RECEIVED"); return; case TCPS_ESTABLISHED: db_printf("TCPS_ESTABLISHED"); return; case TCPS_CLOSE_WAIT: db_printf("TCPS_CLOSE_WAIT"); return; case TCPS_FIN_WAIT_1: db_printf("TCPS_FIN_WAIT_1"); return; case TCPS_CLOSING: db_printf("TCPS_CLOSING"); return; case TCPS_LAST_ACK: db_printf("TCPS_LAST_ACK"); return; case TCPS_FIN_WAIT_2: db_printf("TCPS_FIN_WAIT_2"); return; case TCPS_TIME_WAIT: db_printf("TCPS_TIME_WAIT"); return; default: db_printf("unknown"); return; } } static void db_print_tflags(u_int t_flags) { int comma; comma = 0; if (t_flags & TF_ACKNOW) { db_printf("%sTF_ACKNOW", comma ? ", " : ""); comma = 1; } if (t_flags & TF_DELACK) { db_printf("%sTF_DELACK", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NODELAY) { db_printf("%sTF_NODELAY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NOOPT) { db_printf("%sTF_NOOPT", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SENTFIN) { db_printf("%sTF_SENTFIN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_REQ_SCALE) { db_printf("%sTF_REQ_SCALE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_RCVD_SCALE) { db_printf("%sTF_RECVD_SCALE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_REQ_TSTMP) { db_printf("%sTF_REQ_TSTMP", comma ? ", " : ""); comma = 1; } if (t_flags & TF_RCVD_TSTMP) { db_printf("%sTF_RCVD_TSTMP", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SACK_PERMIT) { db_printf("%sTF_SACK_PERMIT", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NEEDSYN) { db_printf("%sTF_NEEDSYN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NEEDFIN) { db_printf("%sTF_NEEDFIN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_NOPUSH) { db_printf("%sTF_NOPUSH", comma ? ", " : ""); comma = 1; } if (t_flags & TF_PREVVALID) { db_printf("%sTF_PREVVALID", comma ? ", " : ""); comma = 1; } if (t_flags & TF_WAKESOR) { db_printf("%sTF_WAKESOR", comma ? ", " : ""); comma = 1; } if (t_flags & TF_GPUTINPROG) { db_printf("%sTF_GPUTINPROG", comma ? ", " : ""); comma = 1; } if (t_flags & TF_MORETOCOME) { db_printf("%sTF_MORETOCOME", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SONOTCONN) { db_printf("%sTF_SONOTCONN", comma ? ", " : ""); comma = 1; } if (t_flags & TF_LASTIDLE) { db_printf("%sTF_LASTIDLE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_RXWIN0SENT) { db_printf("%sTF_RXWIN0SENT", comma ? ", " : ""); comma = 1; } if (t_flags & TF_FASTRECOVERY) { db_printf("%sTF_FASTRECOVERY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_WASFRECOVERY) { db_printf("%sTF_WASFRECOVERY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_SIGNATURE) { db_printf("%sTF_SIGNATURE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_FORCEDATA) { db_printf("%sTF_FORCEDATA", comma ? ", " : ""); comma = 1; } if (t_flags & TF_TSO) { db_printf("%sTF_TSO", comma ? ", " : ""); comma = 1; } if (t_flags & TF_TOE) { db_printf("%sTF_TOE", comma ? ", " : ""); comma = 1; } if (t_flags & TF_CLOSED) { db_printf("%sTF_CLOSED", comma ? ", " : ""); comma = 1; } if (t_flags & TF_UNUSED1) { db_printf("%sTF_UNUSED1", comma ? ", " : ""); comma = 1; } if (t_flags & TF_LRD) { db_printf("%sTF_LRD", comma ? ", " : ""); comma = 1; } if (t_flags & TF_CONGRECOVERY) { db_printf("%sTF_CONGRECOVERY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_WASCRECOVERY) { db_printf("%sTF_WASCRECOVERY", comma ? ", " : ""); comma = 1; } if (t_flags & TF_FASTOPEN) { db_printf("%sTF_FASTOPEN", comma ? ", " : ""); comma = 1; } } static void db_print_tflags2(u_int t_flags2) { int comma; comma = 0; if (t_flags2 & TF2_PLPMTU_BLACKHOLE) { db_printf("%sTF2_PLPMTU_BLACKHOLE", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_PLPMTU_PMTUD) { db_printf("%sTF2_PLPMTU_PMTUD", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_PLPMTU_MAXSEGSNT) { db_printf("%sTF2_PLPMTU_MAXSEGSNT", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_LOG_AUTO) { db_printf("%sTF2_LOG_AUTO", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_DROP_AF_DATA) { db_printf("%sTF2_DROP_AF_DATA", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_ECN_PERMIT) { db_printf("%sTF2_ECN_PERMIT", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_ECN_SND_CWR) { db_printf("%sTF2_ECN_SND_CWR", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_ECN_SND_ECE) { db_printf("%sTF2_ECN_SND_ECE", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_ACE_PERMIT) { db_printf("%sTF2_ACE_PERMIT", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_HPTS_CPU_SET) { db_printf("%sTF2_HPTS_CPU_SET", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_FBYTES_COMPLETE) { db_printf("%sTF2_FBYTES_COMPLETE", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_ECN_USE_ECT1) { db_printf("%sTF2_ECN_USE_ECT1", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_TCP_ACCOUNTING) { db_printf("%sTF2_TCP_ACCOUNTING", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_HPTS_CALLS) { db_printf("%sTF2_HPTS_CALLS", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_MBUF_L_ACKS) { db_printf("%sTF2_MBUF_L_ACKS", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_MBUF_ACKCMP) { db_printf("%sTF2_MBUF_ACKCMP", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_SUPPORTS_MBUFQ) { db_printf("%sTF2_SUPPORTS_MBUFQ", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_MBUF_QUEUE_READY) { db_printf("%sTF2_MBUF_QUEUE_READY", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_DONT_SACK_QUEUE) { db_printf("%sTF2_DONT_SACK_QUEUE", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_CANNOT_DO_ECN) { db_printf("%sTF2_CANNOT_DO_ECN", comma ? ", " : ""); comma = 1; } if (t_flags2 & TF2_NO_ISS_CHECK) { db_printf("%sTF2_NO_ISS_CHECK", comma ? ", " : ""); comma = 1; } } static void db_print_toobflags(char t_oobflags) { int comma; comma = 0; if (t_oobflags & TCPOOB_HAVEDATA) { db_printf("%sTCPOOB_HAVEDATA", comma ? ", " : ""); comma = 1; } if (t_oobflags & TCPOOB_HADDATA) { db_printf("%sTCPOOB_HADDATA", comma ? ", " : ""); comma = 1; } } static void db_print_bblog_state(int state) { switch (state) { case TCP_LOG_STATE_RATIO_OFF: db_printf("TCP_LOG_STATE_RATIO_OFF"); break; case TCP_LOG_STATE_CLEAR: db_printf("TCP_LOG_STATE_CLEAR"); break; case TCP_LOG_STATE_OFF: db_printf("TCP_LOG_STATE_OFF"); break; case TCP_LOG_STATE_TAIL: db_printf("TCP_LOG_STATE_TAIL"); break; case TCP_LOG_STATE_HEAD: db_printf("TCP_LOG_STATE_HEAD"); break; case TCP_LOG_STATE_HEAD_AUTO: db_printf("TCP_LOG_STATE_HEAD_AUTO"); break; case TCP_LOG_STATE_CONTINUAL: db_printf("TCP_LOG_STATE_CONTINUAL"); break; case TCP_LOG_STATE_TAIL_AUTO: db_printf("TCP_LOG_STATE_TAIL_AUTO"); break; case TCP_LOG_VIA_BBPOINTS: db_printf("TCP_LOG_STATE_BBPOINTS"); break; default: db_printf("UNKNOWN(%d)", state); break; } } static void -db_print_tcpcb(struct tcpcb *tp, const char *name, int indent, bool show_bblog) +db_print_tcpcb(struct tcpcb *tp, const char *name, int indent, bool show_bblog, + bool show_inpcb) { db_print_indent(indent); db_printf("%s at %p\n", name, tp); indent += 2; + if (show_inpcb) + db_print_inpcb(tptoinpcb(tp), "t_inpcb", indent); + db_print_indent(indent); db_printf("t_segq first: %p t_segqlen: %d t_dupacks: %d\n", TAILQ_FIRST(&tp->t_segq), tp->t_segqlen, tp->t_dupacks); db_print_indent(indent); db_printf("t_callout: %p t_timers: %p\n", &tp->t_callout, &tp->t_timers); db_print_indent(indent); db_printf("t_state: %d (", tp->t_state); db_print_tstate(tp->t_state); db_printf(")\n"); db_print_indent(indent); db_printf("t_flags: 0x%x (", tp->t_flags); db_print_tflags(tp->t_flags); db_printf(")\n"); db_print_indent(indent); db_printf("t_flags2: 0x%x (", tp->t_flags2); db_print_tflags2(tp->t_flags2); db_printf(")\n"); db_print_indent(indent); db_printf("snd_una: 0x%08x snd_max: 0x%08x snd_nxt: 0x%08x\n", tp->snd_una, tp->snd_max, tp->snd_nxt); db_print_indent(indent); db_printf("snd_up: 0x%08x snd_wl1: 0x%08x snd_wl2: 0x%08x\n", tp->snd_up, tp->snd_wl1, tp->snd_wl2); db_print_indent(indent); db_printf("iss: 0x%08x irs: 0x%08x rcv_nxt: 0x%08x\n", tp->iss, tp->irs, tp->rcv_nxt); db_print_indent(indent); db_printf("rcv_adv: 0x%08x rcv_wnd: %u rcv_up: 0x%08x\n", tp->rcv_adv, tp->rcv_wnd, tp->rcv_up); db_print_indent(indent); db_printf("snd_wnd: %u snd_cwnd: %u\n", tp->snd_wnd, tp->snd_cwnd); db_print_indent(indent); db_printf("snd_ssthresh: %u snd_recover: " "0x%08x\n", tp->snd_ssthresh, tp->snd_recover); db_print_indent(indent); db_printf("t_rcvtime: %u t_startime: %u\n", tp->t_rcvtime, tp->t_starttime); db_print_indent(indent); db_printf("t_rttime: %u t_rtsq: 0x%08x\n", tp->t_rtttime, tp->t_rtseq); db_print_indent(indent); db_printf("t_rxtcur: %d t_maxseg: %u t_srtt: %d\n", tp->t_rxtcur, tp->t_maxseg, tp->t_srtt); db_print_indent(indent); db_printf("t_rttvar: %d t_rxtshift: %d t_rttmin: %u\n", tp->t_rttvar, tp->t_rxtshift, tp->t_rttmin); db_print_indent(indent); db_printf("t_rttupdated: %u max_sndwnd: %u t_softerror: %d\n", tp->t_rttupdated, tp->max_sndwnd, tp->t_softerror); db_print_indent(indent); db_printf("t_oobflags: 0x%x (", tp->t_oobflags); db_print_toobflags(tp->t_oobflags); db_printf(") t_iobc: 0x%02x\n", tp->t_iobc); db_print_indent(indent); db_printf("snd_scale: %u rcv_scale: %u request_r_scale: %u\n", tp->snd_scale, tp->rcv_scale, tp->request_r_scale); db_print_indent(indent); db_printf("ts_recent: %u ts_recent_age: %u\n", tp->ts_recent, tp->ts_recent_age); db_print_indent(indent); db_printf("ts_offset: %u last_ack_sent: 0x%08x snd_cwnd_prev: " "%u\n", tp->ts_offset, tp->last_ack_sent, tp->snd_cwnd_prev); db_print_indent(indent); db_printf("snd_ssthresh_prev: %u snd_recover_prev: 0x%08x " "t_badrxtwin: %u\n", tp->snd_ssthresh_prev, tp->snd_recover_prev, tp->t_badrxtwin); db_print_indent(indent); db_printf("snd_numholes: %d snd_holes first: %p\n", tp->snd_numholes, TAILQ_FIRST(&tp->snd_holes)); db_print_indent(indent); db_printf("snd_fack: 0x%08x rcv_numsacks: %d\n", tp->snd_fack, tp->rcv_numsacks); /* Skip sackblks, sackhint. */ db_print_indent(indent); db_printf("t_rttlow: %d rfbuf_ts: %u rfbuf_cnt: %d\n", tp->t_rttlow, tp->rfbuf_ts, tp->rfbuf_cnt); db_print_indent(indent); db_printf("t_fb.tfb_tcp_block_name: %s\n", tp->t_fb->tfb_tcp_block_name); db_print_indent(indent); db_printf("t_cc.name: %s\n", tp->t_cc->name); db_print_indent(indent); db_printf("_t_logstate: %d (", tp->_t_logstate); db_print_bblog_state(tp->_t_logstate); db_printf(")\n"); db_print_indent(indent); db_printf("t_lognum: %d t_loglimit: %d t_logsn: %u\n", tp->t_lognum, tp->t_loglimit, tp->t_logsn); if (show_bblog) { #ifdef TCP_BLACKBOX db_print_bblog_entries(&tp->t_logs, indent); #else db_print_indent(indent); db_printf("BBLog not supported\n"); #endif } } DB_SHOW_COMMAND(tcpcb, db_show_tcpcb) { struct tcpcb *tp; - bool show_bblog; + bool show_bblog, show_inpcb; if (!have_addr) { - db_printf("usage: show tcpcb \n"); + db_printf("usage: show tcpcb[/bi] \n"); return; } show_bblog = strchr(modif, 'b') != NULL; + show_inpcb = strchr(modif, 'i') != NULL; tp = (struct tcpcb *)addr; - - db_print_tcpcb(tp, "tcpcb", 0, show_bblog); + db_print_tcpcb(tp, "tcpcb", 0, show_bblog, show_inpcb); } DB_SHOW_ALL_COMMAND(tcpcbs, db_show_all_tcpcbs) { VNET_ITERATOR_DECL(vnet_iter); struct inpcb *inp; - bool only_locked, show_bblog; + struct tcpcb *tp; + bool only_locked, show_bblog, show_inpcb; only_locked = strchr(modif, 'l') != NULL; show_bblog = strchr(modif, 'b') != NULL; + show_inpcb = strchr(modif, 'i') != NULL; VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); CK_LIST_FOREACH(inp, &V_tcbinfo.ipi_listhead, inp_list) { if (only_locked && inp->inp_lock.rw_lock == RW_UNLOCKED) continue; - db_print_tcpcb(intotcpcb(inp), "tcpcb", 0, show_bblog); + tp = intotcpcb(inp); + db_print_tcpcb(tp, "tcpcb", 0, show_bblog, show_inpcb); if (db_pager_quit) break; } CURVNET_RESTORE(); if (db_pager_quit) break; } } #endif