/* iobuf.h - I/O buffer * Copyright (C) 1998, 1999, 2000, 2001, 2003, * 2010 Free Software Foundation, Inc. * * This file is part of GnuPG. * * This file is free software; you can redistribute it and/or modify * it under the terms of either * * - the GNU Lesser General Public License as published by the Free * Software Foundation; either version 3 of the License, or (at * your option) any later version. * * or * * - the GNU General Public License as published by the Free * Software Foundation; either version 2 of the License, or (at * your option) any later version. * * or both in parallel, as here. * * This file is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, see . */ #ifndef GNUPG_COMMON_IOBUF_H #define GNUPG_COMMON_IOBUF_H /* An iobuf is basically a filter in a pipeline. Consider the following command, which consists of three filters that are chained together: $ cat file | base64 --decode | gunzip The first filter reads the file from the file system and sends that data to the second filter. The second filter decodes base64-encoded data and sends the data to the third and last filter. The last filter decompresses the data and the result is displayed on the terminal. The iobuf system works in the same way where each iobuf is a filter and the individual iobufs can be chained together. There are number of predefined filters. iobuf_open(), for instance, creates a filter that reads from a specified file. And, iobuf_temp_with_content() creates a filter that returns some specified contents. There are also filters for writing content. iobuf_openrw opens a file for writing. iobuf_temp creates a filter that writes data to a fixed-sized buffer. To chain filters together, you use the iobuf_push_filter() function. The filters are chained together using the chain field in the iobuf_t. A pipeline can only be used for reading (IOBUF_INPUT) or for writing (IOBUF_OUTPUT / IOBUF_OUTPUT_TEMP). When reading, data flows from the last filter towards the first. That is, the user calls iobuf_read(), the module reads from the first filter, which gets its input from the second filter, etc. When writing, data flows from the first filter towards the last. In this case, when the user calls iobuf_write(), the data is written to the first filter, which writes the transformed data to the second filter, etc. An iobuf_t contains some state about the filter. For instance, it indicates if the filter has already returned EOF (filter_eof) and the next filter in the pipeline, if any (chain). It also contains a function pointer, filter. This is a generic function. It is called when input is needed or output is available. In this case it is passed a pointer to some filter-specific persistent state (filter_ov), the actual operation, the next filter in the chain, if any, and a buffer that either contains the contents to write, if the pipeline is setup to write data, or is the place to store data, if the pipeline is setup to read data. Unlike a Unix pipeline, an IOBUF pipeline can return EOF multiple times. This is similar to the following: { cat file1; cat file2; } | grep foo However, instead of grep seeing a single stream, grep would see each byte stream followed by an EOF marker. (When a filter returns EOF, the EOF is returned to the user exactly once and then the filter is removed from the pipeline.) */ /* For estream_t. */ #include #include "../common/types.h" #include "../common/sysutils.h" #define DBG_IOBUF iobuf_debug_mode /* Filter control modes. */ enum { IOBUFCTRL_INIT = 1, IOBUFCTRL_FREE = 2, IOBUFCTRL_UNDERFLOW = 3, IOBUFCTRL_FLUSH = 4, IOBUFCTRL_DESC = 5, IOBUFCTRL_CANCEL = 6, IOBUFCTRL_USER = 16 }; /* Command codes for iobuf_ioctl. */ typedef enum { IOBUF_IOCTL_KEEP_OPEN = 1, /* Uses intval. */ IOBUF_IOCTL_INVALIDATE_CACHE = 2, /* Uses ptrval. */ IOBUF_IOCTL_NO_CACHE = 3, /* Uses intval. */ IOBUF_IOCTL_FSYNC = 4 /* Uses ptrval. */ } iobuf_ioctl_t; enum iobuf_use { /* Pipeline is in input mode. The data flows from the end to the beginning. That is, when reading from the pipeline, the first filter gets its input from the second filter, etc. */ IOBUF_INPUT, /* Pipeline is in input mode. The last filter in the pipeline is a temporary buffer from which the data is "read". */ IOBUF_INPUT_TEMP, /* Pipeline is in output mode. The data flows from the beginning to the end. That is, when writing to the pipeline, the user writes to the first filter, which transforms the data and sends it to the second filter, etc. */ IOBUF_OUTPUT, /* Pipeline is in output mode. The last filter in the pipeline is a temporary buffer that grows as necessary. */ IOBUF_OUTPUT_TEMP }; typedef struct iobuf_struct *iobuf_t; typedef struct iobuf_struct *IOBUF; /* Compatibility with gpg 1.4. */ /* fixme: we should hide most of this stuff */ struct iobuf_struct { /* The type of filter. Either IOBUF_INPUT, IOBUF_OUTPUT or IOBUF_OUTPUT_TEMP. */ enum iobuf_use use; /* nlimit can be changed using iobuf_set_limit. If non-zero, it is the number of additional bytes that can be read from the filter before EOF is forcefully returned. */ off_t nlimit; /* nbytes if the number of bytes that have been read (using iobuf_get / iobuf_readbyte / iobuf_read) since the last call to iobuf_set_limit. */ off_t nbytes; /* The number of bytes read prior to the last call to iobuf_set_limit. Thus, the total bytes read (i.e., the position of stream) is ntotal + nbytes. */ off_t ntotal; /* Whether we need to read from the filter one byte at a time or whether we can do bulk reads. We need to read one byte at a time if a limit (set via iobuf_set_limit) is active. */ int nofast; /* A buffer for unread/unwritten data. For an output pipeline (IOBUF_OUTPUT), this is the data that has not yet been written to the filter. Consider a simple pipeline consisting of a single stage, which writes to a file. When you write to the pipeline (iobuf_writebyte or iobuf_write), the data is first stored in this buffer. Only when the buffer is full or you call iobuf_flush() is FILTER actually called and the data written to the file. For an input pipeline (IOBUF_INPUT), this is the data that has been read from this filter, but not yet been read from the preceding filter (or the user, if this filter is the head of the pipeline). Again, consider a simple pipeline consisting of a single stage. This stage reads from a file. If you read a single byte (iobuf_get) and the buffer is empty, then FILTER is called to fill the buffer. In this case, a single byte is not requested, but the whole buffer is filled (if possible). */ struct { /* Size of the buffer. */ size_t size; /* Number of bytes at the beginning of the buffer that have already been consumed. (In other words: the index of the first byte that hasn't been consumed.) This is only non-zero for input filters. */ size_t start; /* The number of bytes in the buffer including any bytes that have been consumed. */ size_t len; /* The buffer itself. */ byte *buf; } d; /* When FILTER is called to read some data, it may read some data and then return EOF. We can't return the EOF immediately. Instead, we note that we observed the EOF and when the buffer is finally empty, we return the EOF. */ int filter_eof; /* Like filter_eof, when FILTER is called to read some data, it may read some data and then return an error. We can't return the error (in the form of an EOF) immediately. Instead, we note that we observed the error and when the buffer is finally empty, we return the EOF. */ int error; /* The callback function to read data from the filter, etc. See iobuf_filter_push for details. */ int (*filter) (void *opaque, int control, iobuf_t chain, byte * buf, size_t * len); /* An opaque pointer that can be used for local filter state. This is passed as the first parameter to FILTER. */ void *filter_ov; /* Whether the iobuf code should free(filter_ov) when destroying the filter. */ int filter_ov_owner; /* When using iobuf_open, iobuf_create, iobuf_openrw to open a file, the file's name is saved here. This is used to delete the file when an output pipeline (IOBUF_OUPUT) is canceled (iobuf_cancel). */ char *real_fname; /* The next filter in the pipeline. */ iobuf_t chain; /* This field is for debugging. Each time a filter is allocated (via iobuf_alloc()), a monotonically increasing counter is incremented and this field is set to the new value. This field should only be accessed via the iobuf_io macro. */ int no; /* The number of filters in the pipeline following (not including) this one. When you call iobuf_push_filter or iobuf_push_filter2, this value is used to check the length of the pipeline if the pipeline already contains 65 stages then these functions fail. This amount of nesting typically indicates corrupted data or an active denial of service attack. */ int subno; }; extern int iobuf_debug_mode; /* Returns whether the specified filename corresponds to a pipe. In particular, this function checks if FNAME is "-" and, if special filenames are enabled (see check_special_filename), whether FNAME is a special filename. */ int iobuf_is_pipe_filename (const char *fname); /* Allocate a new filter. This filter doesn't have a function assigned to it. Thus you need to manually set IOBUF->FILTER and IOBUF->FILTER_OV, if required. This function is intended to help create a new primary source or primary sink, i.e., the last filter in the pipeline. USE is IOBUF_INPUT, IOBUF_INPUT_TEMP, IOBUF_OUTPUT or IOBUF_OUTPUT_TEMP. BUFSIZE is the desired internal buffer size (that is, the size of the typical read / write request). */ iobuf_t iobuf_alloc (int use, size_t bufsize); /* Create an output filter that simply buffers data written to it. This is useful for collecting data for later processing. The buffer can be written to in the usual way (iobuf_write, etc.). The data can later be extracted using iobuf_write_temp() or iobuf_temp_to_buffer(). */ iobuf_t iobuf_temp (void); /* Create an input filter that contains some data for reading. */ iobuf_t iobuf_temp_with_content (const char *buffer, size_t length); /* Create an input file filter that reads from a file. If FNAME is '-', reads from stdin. If special filenames are enabled (iobuf_enable_special_filenames), then interprets special filenames. */ iobuf_t iobuf_open (const char *fname); /* Create an output file filter that writes to a file. If FNAME is NULL or '-', writes to stdout. If special filenames are enabled (iobuf_enable_special_filenames), then interprets special filenames. If FNAME is not NULL, '-' or a special filename, the file is opened for writing. If the file exists, it is truncated. If MODE700 is TRUE, the file is created with mode 600. Otherwise, mode 666 is used. */ iobuf_t iobuf_create (const char *fname, int mode700); /* Create an output file filter that writes to a specified file. Neither '-' nor special file names are recognized. */ iobuf_t iobuf_openrw (const char *fname); /* Create a file filter using an existing file descriptor. If MODE contains the letter 'w', creates an output filter. Otherwise, creates an input filter. Note: MODE must reflect the file descriptors actual mode! When the filter is destroyed, the file descriptor is closed. */ iobuf_t iobuf_fdopen (int fd, const char *mode); /* Like iobuf_fdopen, but doesn't close the file descriptor when the filter is destroyed. */ iobuf_t iobuf_fdopen_nc (int fd, const char *mode); /* Create a filter using an existing estream. If MODE contains the letter 'w', creates an output filter. Otherwise, creates an input filter. If KEEP_OPEN is TRUE, then the stream is not closed when the filter is destroyed. Otherwise, the stream is closed when the filter is destroyed. */ iobuf_t iobuf_esopen (estream_t estream, const char *mode, int keep_open); /* Create a filter using an existing socket. On Windows creates a special socket filter. On non-Windows systems simply, this simply calls iobuf_fdopen. */ iobuf_t iobuf_sockopen (int fd, const char *mode); /* Set various options / perform different actions on a PIPELINE. See the IOBUF_IOCTL_* macros above. */ int iobuf_ioctl (iobuf_t a, iobuf_ioctl_t cmd, int intval, void *ptrval); /* Close a pipeline. The filters in the pipeline are first flushed using iobuf_flush, if they are output filters, and then IOBUFCTRL_FREE is called on each filter. If any filter returns a non-zero value in response to the IOBUFCTRL_FREE, that first such non-zero value is returned. Note: processing is not aborted in this case. If all filters are freed successfully, 0 is returned. */ int iobuf_close (iobuf_t iobuf); /* Calls IOBUFCTRL_CANCEL on each filter in the pipeline. Then calls io_close() on the pipeline. Finally, if the pipeline is an output pipeline, deletes the file. Returns the result of calling iobuf_close on the pipeline. */ int iobuf_cancel (iobuf_t iobuf); /* Add a new filter to the front of a pipeline. A is the head of the pipeline. F is the filter implementation. OV is an opaque pointer that is passed to F and is normally used to hold any internal state, such as a file pointer. Note: you may only maintain a reference to an iobuf_t as a reference to the head of the pipeline. That is, don't think about setting a pointer in OV to point to the filter's iobuf_t. This is because when we add a new filter to a pipeline, we memcpy the state in A into new buffer. This has the advantage that there is no need to update any references to the pipeline when a filter is added or removed, but it also means that a filter's state moves around in memory. The behavior of the filter function is determined by the value of the control parameter: IOBUFCTRL_INIT: Called this value just before the filter is linked into the pipeline. This can be used to initialize internal data structures. IOBUFCTRL_FREE: Called with this value just before the filter is removed from the pipeline. Normally used to release internal data structures, close a file handle, etc. IOBUFCTRL_UNDERFLOW: Called with this value to fill the passed buffer with more data. *LEN is the size of the buffer. Before returning, it should be set to the number of bytes which were written into the buffer. The function must return 0 to indicate success, -1 on EOF and a GPG_ERR_xxxxx code for any error. Note: this function may both return data and indicate an error or EOF. In this case, it simply writes the data to BUF, sets *LEN and returns the appropriate return code. The implication is that if an error occurs and no data has yet been written, it is essential that *LEN be set to 0! IOBUFCTRL_FLUSH: Called with this value to write out any collected data. *LEN is the number of bytes in BUF that need to be written out. Returns 0 on success and a GPG_ERR_* code otherwise. *LEN must be set to the number of bytes that were written out. IOBUFCTRL_CANCEL: Called with this value when iobuf_cancel() is called on the pipeline. IOBUFCTRL_DESC: Called with this value to get a human-readable description of the filter. *LEN is the size of the buffer. The description is filled into BUF, NUL-terminated. Always returns 0. */ int iobuf_push_filter (iobuf_t a, int (*f) (void *opaque, int control, iobuf_t chain, byte * buf, size_t * len), void *ov); /* This variant of iobuf_push_filter allows the called to indicate that OV should be freed when this filter is freed. That is, if REL_OV is TRUE, then when the filter is popped or freed OV will be freed after the filter function is called with control set to IOBUFCTRL_FREE. */ int iobuf_push_filter2 (iobuf_t a, int (*f) (void *opaque, int control, iobuf_t chain, byte * buf, size_t * len), void *ov, int rel_ov); /* Pop the top filter. The top filter must have the filter function F and the cookie OV. The cookie check is ignored if OV is NULL. */ int iobuf_pop_filter (iobuf_t a, int (*f) (void *opaque, int control, iobuf_t chain, byte * buf, size_t * len), void *ov); /* Used for debugging. Prints out the chain using log_debug if IOBUF_DEBUG_MODE is not 0. */ int iobuf_print_chain (iobuf_t a); /* Indicate that some error occurred on the specified filter. */ #define iobuf_set_error(a) do { (a)->error = 1; } while(0) /* Return any pending error on filter A. */ #define iobuf_error(a) ((a)->error) /* Limit the amount of additional data that may be read from the filter. That is, if you've already read 100 bytes from A and you set the limit to 50, then you can read up to an additional 50 bytes (i.e., a total of 150 bytes) before EOF is forcefully returned. Setting NLIMIT to 0 removes any active limit. Note: using iobuf_seek removes any currently enforced limit! */ void iobuf_set_limit (iobuf_t a, off_t nlimit); /* Returns the number of bytes that have been read from the pipeline. Note: the result is undefined for IOBUF_OUTPUT and IOBUF_OUTPUT_TEMP pipelines! */ off_t iobuf_tell (iobuf_t a); /* There are two cases: - If A is an INPUT or OUTPUT pipeline, then the last filter in the pipeline is found. If that is not a file filter, -1 is returned. Otherwise, an fseek(..., SEEK_SET) is performed on the file descriptor. - If A is a TEMP pipeline and the *first* (and thus only filter) is a TEMP filter, then the "file position" is effectively unchanged. That is, data is appended to the buffer and the seek does not cause the size of the buffer to grow. If no error occurred, then any limit previous set by iobuf_set_limit() is cleared. Further, any error on the filter (the file filter or the temp filter) is cleared. Returns 0 on success and -1 if an error occurs. */ int iobuf_seek (iobuf_t a, off_t newpos); /* Read a single byte. If a filter has no more data, returns -1 to indicate the EOF. Generally, you don't want to use this function, but instead prefer the iobuf_get macro, which is faster if there is data in the internal buffer. */ int iobuf_readbyte (iobuf_t a); /* Get a byte from the iobuf; must check for eof prior to this function. This function returns values in the range 0 .. 255 or -1 to indicate EOF. iobuf_get_noeof() does not return -1 to indicate EOF, but masks the returned value to be in the range 0 .. 255. */ #define iobuf_get(a) \ ( ((a)->nofast || (a)->d.start >= (a)->d.len )? \ iobuf_readbyte((a)) : ( (a)->nbytes++, (a)->d.buf[(a)->d.start++] ) ) #define iobuf_get_noeof(a) (iobuf_get((a))&0xff) /* Fill BUF with up to BUFLEN bytes. If a filter has no more data, returns -1 to indicate the EOF. Otherwise returns the number of bytes read. */ int iobuf_read (iobuf_t a, void *buf, unsigned buflen); /* Read a line of input (including the '\n') from the pipeline. The semantics are the same as for fgets(), but if the buffer is too short a larger one will be allocated up to *MAX_LENGTH and the end of the line except the trailing '\n' discarded. (Thus, *ADDR_OF_BUFFER must be allocated using malloc().) If the buffer is enlarged, then *LENGTH_OF_BUFFER will be updated to reflect the new size. If the line is truncated, then *MAX_LENGTH will be set to 0. If *ADDR_OF_BUFFER is NULL, a buffer is allocated using malloc(). A line is considered a byte stream ending in a '\n'. Returns the number of characters written to the buffer (i.e., excluding any discarded characters due to truncation). Thus, use this instead of strlen(buffer) to determine the length of the string as this is unreliable if the input contains NUL characters. EOF is indicated by a line of length zero. The last LF may be missing due to an EOF. */ unsigned iobuf_read_line (iobuf_t a, byte ** addr_of_buffer, unsigned *length_of_buffer, unsigned *max_length); /* Read up to BUFLEN bytes from pipeline A. Note: this function can't return more than the pipeline's internal buffer size. The return value is the number of bytes actually written to BUF. If the filter returns EOF, then this function returns -1. This function does not clear any pending EOF. That is, if the pipeline consists of two filters and the first one returns EOF during the peek, then the subsequent iobuf_read* will still return EOF before returning the data from the second filter. */ int iobuf_peek (iobuf_t a, byte * buf, unsigned buflen); /* Write a byte to the pipeline. Returns 0 on success and an error code otherwise. */ int iobuf_writebyte (iobuf_t a, unsigned c); /* Alias for iobuf_writebyte. */ #define iobuf_put(a,c) iobuf_writebyte(a,c) /* Write a sequence of bytes to the pipeline. Returns 0 on success and an error code otherwise. */ int iobuf_write (iobuf_t a, const void *buf, unsigned buflen); /* Write a string (not including the NUL terminator) to the pipeline. Returns 0 on success and an error code otherwise. */ int iobuf_writestr (iobuf_t a, const char *buf); /* Flushes the pipeline removing all filters but the sink (the last filter) in the process. */ void iobuf_flush_temp (iobuf_t temp); /* Flushes the pipeline SOURCE removing all filters but the sink (the last filter) in the process (i.e., it calls iobuf_flush_temp(source)) and then writes the data to the pipeline DEST. Note: this doesn't free (iobuf_close()) SOURCE. Both SOURCE and DEST must be output pipelines. */ int iobuf_write_temp (iobuf_t dest, iobuf_t source); /* Flushes each filter in the pipeline (i.e., sends any buffered data to the filter by calling IOBUFCTRL_FLUSH). Then, copies up to the first BUFLEN bytes from the last filter's internal buffer (which will only be non-empty if it is a temp filter) to the buffer BUFFER. Returns the number of bytes actually copied. */ size_t iobuf_temp_to_buffer (iobuf_t a, byte * buffer, size_t buflen); /* Copies the data from the input iobuf SOURCE to the output iobuf DEST until either an error is encountered or EOF is reached. Returns the number of bytes successfully written. If an error occurred, then any buffered bytes are not returned to SOURCE and are effectively lost. To check if an error occurred, use iobuf_error. */ size_t iobuf_copy (iobuf_t dest, iobuf_t source); /* Return the size of any underlying file. This only works with file_filter based pipelines. On Win32, it is sometimes not possible to determine the size of files larger than 4GB. In this case, *OVERFLOW (if not NULL) is set to 1. Otherwise, *OVERFLOW is set to 0. */ off_t iobuf_get_filelength (iobuf_t a, int *overflow); #define IOBUF_FILELENGTH_LIMIT 0xffffffff /* Return the file descriptor designating the underlying file. This only works with file_filter based pipelines. */ int iobuf_get_fd (iobuf_t a); /* Return the real filename, if available. This only supports pipelines that end in file filters. Returns NULL if not available. */ const char *iobuf_get_real_fname (iobuf_t a); /* Return the filename or a description thereof. For instance, for iobuf_open("-"), this will return "[stdin]". This only supports pipelines that end in file filters. Returns NULL if not available. */ const char *iobuf_get_fname (iobuf_t a); /* Like iobuf_getfname, but instead of returning NULL if no description is available, return "[?]". */ const char *iobuf_get_fname_nonnull (iobuf_t a); /* Pushes a filter on the pipeline that interprets the datastream as an OpenPGP data block whose length is encoded using partial body length headers (see Section 4.2.2.4 of RFC 4880). Concretely, it just returns / writes the data and finishes the packet with an EOF. */ void iobuf_set_partial_body_length_mode (iobuf_t a, size_t len); /* If PARTIAL is set, then read from the pipeline until the first EOF is returned. If PARTIAL is 0, then read up to N bytes or until the first EOF is returned. Recall: a filter can return EOF. In this case, it and all preceding filters are popped from the pipeline and the next read is from the following filter (which may or may not return EOF). */ void iobuf_skip_rest (iobuf_t a, unsigned long n, int partial); #define iobuf_where(a) "[don't know]" /* Each time a filter is allocated (via iobuf_alloc()), a monotonically increasing counter is incremented and this field is set to the new value. This macro returns that number. */ #define iobuf_id(a) ((a)->no) #define iobuf_get_temp_buffer(a) ( (a)->d.buf ) #define iobuf_get_temp_length(a) ( (a)->d.len ) /* Whether the filter uses an in-memory buffer. */ #define iobuf_is_temp(a) ( (a)->use == IOBUF_OUTPUT_TEMP ) #endif /*GNUPG_COMMON_IOBUF_H*/