< Previous | |
Product: Volume Manager Manual Pages for Storage Foundation | |
Manual: Miscellaneous (7) |
vxtraceNAMEvxtrace - VERITAS Volume Manager I/O Tracing Device SYNOPSIS/dev/vx/trace DESCRIPTIONThe vxtrace device implements the VERITAS Volume Manager (VxVM) I/O tracing and the error tracing. An I/O tracing interface is available that users or processes can use to get a trace of I/Os for specified sets of kernel objects. Each separate user of the I/O tracing interface can specify the set of desired trace data independent of all other users. I/O events include regular read and write operations, special I/O operations (ioctls), as well as special recovery operations (for example, recovery reads). A special tracing mechanism exists for getting error trace data. The error tracing mechanism is independent of any I/O tracing and is always enabled for all pertinent kernel I/O objects. It is possible for a process to get both a set of saved errors and to wait for new errors. IOCTLSThe format for calling each ioctl command is: #include <sys/types.h> #include <vxvm/voltrace.h> struct tag arg; int ioctl (int fd, int cmd, struct tag arg); The first argument fd is a file descriptor which is returned from opening the /dev/vx/trace device. Each tracing device opened is a cloned device which can be used as a private kernel trace channel. The value of cmd is the ioctl command code, and arg is usually a pointer to a structure containing the arguments that need to be passed to the kernel. The return value for all these ioctls is 0 if the command was successful, and -1 if it was rejected. If the return value is -1, errno is set to indicate the cause of the error. The following ioctl commands are supported: This command accepts no argument. The VOLIOT_ERROR_TRACE ioctl initializes a kernel trace channel to return error trace data. The trace channel will be initialized to return any previously accumulated error trace data that has not yet been discarded. The accumulated trace data can be skipped by issuing VOLIOT_DISCARD on the channel. This call can be issued on a trace channel that was previously initialized either for error tracing or for regular I/O tracing. In this case, the channel is effectively closed down and then reinitialized as described above. To get the error trace data, issue the read(2) system call. The error trace data consists of a set of variable length trace event records. The first byte of each record indicates the length, in bytes, of the entire record (including the length byte), the second byte indicates the type of the entry (which can be used to determine the format of the entry). Each call to read() returns an integral number of trace event records, not to exceed the number of bytes requested in the read() call; the return value from read() will be adjusted to the number of bytes of trace data actually returned. If the O_NONBLOCK flag is set on the trace channel, and no trace data is available, EAGAIN will be returned; otherwise, the read will block interruptibly until at least one trace record is available. When some trace data is available, the available unread trace records will be returned, up to the limit specified in the call to read(). If more trace records are available, subsequent reads will return those records. The VOLIOT_IO_TRACE_INIT ioctl initializes a kernel trace channel to return I/O trace data. This command accepts bufsize as the argument. Initially, no objects are selected for I/O tracing. To select objects to trace, issue the VOLIOT_IO_TRACE ioctl. The bufsize argument specifies the kernel buffer size to use for gathering events. A larger size reduces the chance that events are lost due to scheduling delays in the event reading process. A bufsize value of 0 requests a default size which is considered reasonable for the system. The value of bufsize will be silently truncated to a maximum value to avoid extreme use of system resources. A bufsize value of (size_t)-1 will yield the maximum buffer size. VOLIOT_IO_TRACE,VOLIOT_IO_UNTRACE The VOLIOT_IO_TRACE and VOLIOT_IO_UNTRACE ioctls enable and disable, respectively, I/O tracing for particular sets of objects on an I/O tracing channel. They both accept a voliot_want_list structure tracelist as the argument. The tracelist argument specifies object sets. The voliot_want_list structure specifies an array of desired object sets. Each object set is identified by a union of structures (the voliot_want_set union), each representing different types of object sets. See the declaration of these structures in voltrace.h for more detail. FILESThe VOLIOT_IO_TRACE and VOLIOT_IO_UNTRACE ioctls enable and disable, respectively, I/O tracing for particular sets of objects on an I/O tracing channel. They both accept a voliot_want_list structure tracelist as the argument. The tracelist argument specifies object sets. The voliot_want_list structure specifies an array of desired object sets. Each object set is identified by a union of structures (the voliot_want_set union), each representing different types of object sets. See the declaration of these structures in voltrace.h for more detail. SEE ALSOvxintro (1M), vxtrace (1M), vxvol (1M) ioctl(2), read(2), vxconfig (7), vxiod (7) |
^ Return to Top | < Previous |
Product: Volume Manager Manual Pages for Storage Foundation | |
Manual: Miscellaneous (7) | |
VERITAS Software Corporation
www.veritas.com |