--- name: "omirr" section: "4" platform: "linux" package: "omirr" version: "0.3-6" release: "Debian 2.2 (Potato)" platforms: ["linux"] url: "https://new.carta.tech/omirr" --- # omirr(4) online mirror kernel log device ## NAME omirr - online mirror kernel log device ## SYNOPSIS **/proc/omirr** \#include \ **int open("/proc/omirr",** **int** *oflag***);** **ssize\_t read(int** *fildes***,** **void \****buf***,** **size\_t** *nbyte***);** **int ioctl(int** *fildes***,** **int** *request***,** **/\*** *arg* **\*/ ...);** ## AVAILABILITY Linux ## DESCRIPTION **/proc/omirr** is a special device in the Linux /proc filesystem used by the [**in.omirrd**(8)](/in.omirrd/8.md) daemon to receive notifications about file system modifications. **/proc/omirr** supports the system calls [**open**(2)](/open/2.md), [**read**(2)](/read/2.md), [**select**(2)](/select/2.md), [**ioctl**(2)](/ioctl/2.md) and [**close**(2)](/close/2.md). The **/proc/omirr** device may be opened only once concurrently. Therefore a second attempt to open the device blocks the calling process unless **O\_NONBLOCK** is specified in *oflags.* The device can only be opened in read-only mode, so **O\_RDONLY** must be passed in the *oflags.* **open()** accepts the following omirr-specific *oflags:* - **O\_NONBLOCK** An open will return without delay even if **/proc/omirr** is already open. In addition, this flag may affect subsequent reads (see below). - **O\_TRUNC** The data already available to be read will be discarded. The behavior of **read()** is also modified by the **O\_NONBLOCK** flag. If set, either via [**open**(2)](/open/2.md) or by a [**fcntl**(2)](/fcntl/2.md) call, and there is no data available, a read attempt returns **-1** with errno set to **EAGAIN.** If not set, the **read()** call blocks until data is available. When data is available, **read()** succeeds and returns the number of bytes actually read and placed in the buffer; this number may be less than the number of bytes requested in the **read()** call. **/proc/omirr** defines the following device-specific **ioctl()** calls, which are defined in **\**: - **OMIRRCLEARBUF** discards the data which is currently available for reading. The **ioctl** *arg* is ignored. - **OMIRRGETBLOCKMODE** returns the current value of the block-mode. The **ioctl** *arg* is of type **(unsigned char \*)**. It is used to return the block-mode value. If block-mode is set, the ring buffer holding the log data cannot overflow, because the process generating the new notification is blocked. If block-mode is not set, the ring-buffer overflow will be detected and logged with a special protocol operation. Block-mode is unset by default. - **OMIRRSETBLOCKMODE** sets block-mode to the value of *arg*, of type **(unsigned char)**. See **OMIRRGETBLOCKMODE,** above, for a description of block-mode. - **OMIRRGETBUFCOUNT** returns the number of bytes available for read. *arg* is of type **(unsigned long \*)**. - **OMIRRGETDONTLOGMASK** returns the current value of the don't-log-mask. The **ioctl** *arg* is of type **(unsigned long \*)**. It is used to return the mask value. The following bits are defined in the mask: - **OMIRR\_DONTLOG\_PROC** suppress logging of file operations by the process which opened **/proc/omirr** - **OMIRR\_DONTLOG\_PPROC** suppress logging of file operations by the parent of the process which opened **/proc/omirr** - **OMIRR\_DONTLOG\_PGRP** suppress logging of file operations by the processes which are in the process group of the process which opened **/proc/omirr** - **OMIRRSETDONTLOGMASK** sets the don't-log-mask to the value of *arg*, of type **(unsigned long)**. See **OMIRRGETDONTLOGMASK,** above, for a description of the don't-log-mask. ## RETURN VALUES **open(),** **read(),** **select(),** **ioctl()** and **close()** return the values described in the manual pages [**open**(2)](/open/2.md), [**read**(2)](/read/2.md), [**select**(2)](/select/2.md), [**ioctl**(2)](/ioctl/2.md) and [**close**(2)](/close/2.md), respectively. Upon encountering an error, **-1** is returned with the ***errno*** variable set accordingly. ## ERRORS In addition to the conditions listed in [**open**(2)](/open/2.md), **open()** can fail if one of the following conditions specific to the /proc/omirr device is met: - **EINVAL** An open mode of **O\_WRONLY** or **O\_RDWR** was specified. - **EWOULDBLOCK** **O\_NONBLOCK** is set and the **/proc/omirr** is already open. - **EINTR** A signal was caught during the **open()** function. In addition to the conditions listed in [**read**(2)](/read/2.md), **read()** can fail if one of the following conditions specific to the /proc/omirr device is met: - **EAGAIN** Non-blocking I/O has been selected using **O\_NONBLOCK** and no data was immediately available for reading from **/proc/omirr**. - **EINTR** A signal was caught during the read operation and no data was transferred. **ioctl()** can fail on the following conditions special to the **/proc/omirr** device: - **EINVAL** *request* or *arg* is not valid for this device. ## FILES - /proc/omirr omirr special proc filesystem device. ## SEE ALSO [**in.omirrd**(8)](/in.omirrd/8.md), [**open**(2)](/open/2.md), [**read**(2)](/read/2.md), [**select**(2)](/select/2.md), [**ioctl**(2)](/ioctl/2.md), [**close**(2)](/close/2.md), [**fcntl**(2)](/fcntl/2.md) ## DIAGNOSTICS - omirr\_log: buffer full - sleeping is an information message generated if block-mode is set and a process generating a new notification is put to sleep, because the notification does not fit into the ring buffer. omirr: setting block mode without OMIRR\_DONTLOG\_PROC may result in dead lock is a warning message generated if block-mode is set using [**ioctl**(2)](/ioctl/2.md) without setting the don't-log-mask to include **OMIRR\_DONTLOG\_PROC** before. See **WARNINGS,** below, for an explanation. ## WARNINGS It is dangerous to switch on block-mode without including **OMIRR\_DONTLOG\_PROC** in the don't-log-mask, because that may result in a dead-lock condition where the process that opened **/proc/omirr** is blocked in a file system operation because the ring buffer is full, and is therefore unable to read data from the ring buffer. To warn about this condition, the kernel generates a warning message, see **DIAGNOSTICS,** above. ## AUTHOR Andreas Luik \