diff options
author | Werner Koch <[email protected]> | 2017-02-16 13:07:27 +0000 |
---|---|---|
committer | Werner Koch <[email protected]> | 2017-02-16 13:07:27 +0000 |
commit | 04bfa6fe6597b8ffcec61cbcacdc7eb137444e80 (patch) | |
tree | 6d966f9e1e27416e0f560cc999825260f32f53cd /common/ksba-io-support.c | |
parent | dirmngr: Prepare certcache for forthcoming changes. (diff) | |
download | gnupg-04bfa6fe6597b8ffcec61cbcacdc7eb137444e80.tar.gz gnupg-04bfa6fe6597b8ffcec61cbcacdc7eb137444e80.zip |
sm,common: Move ksba reader and writer support to common/.
* sm/base64.c: Rename to ...
* common/ksba-io-support.c: this.
* common/ksba-io-support.h: New.
* common/Makefile.am (common_sources): Add new files.
* sm/Makefile.am (gpgsm_SOURCES): Remove base64.c
Signed-off-by: Werner Koch <[email protected]>
Diffstat (limited to 'common/ksba-io-support.c')
-rw-r--r-- | common/ksba-io-support.c | 700 |
1 files changed, 700 insertions, 0 deletions
diff --git a/common/ksba-io-support.c b/common/ksba-io-support.c new file mode 100644 index 000000000..a47e47bf3 --- /dev/null +++ b/common/ksba-io-support.c @@ -0,0 +1,700 @@ +/* kska-io-support.c - Supporting functions for ksba reader and writer + * Copyright (C) 2001, 2003, 2010 Free Software Foundation, Inc. + * + * This file is part of GnuPG. + * + * GnuPG is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation; either version 3 of the License, or + * (at your option) any later version. + * + * GnuPG 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 <https://www.gnu.org/licenses/>. + */ + +#include <config.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <errno.h> +#include <unistd.h> +#include <time.h> +#include <assert.h> + +#include "../sm/gpgsm.h" + + +#include <ksba.h> + +#include "i18n.h" + +#ifdef HAVE_DOSISH_SYSTEM + #define LF "\r\n" +#else + #define LF "\n" +#endif + +/* Data used by the reader callbacks. */ +struct reader_cb_parm_s +{ + estream_t fp; + + unsigned char line[1024]; + int linelen; + int readpos; + int have_lf; + unsigned long line_counter; + + int allow_multi_pem; /* Allow processing of multiple PEM objects. */ + int autodetect; /* Try to detect the input encoding. */ + int assume_pem; /* Assume input encoding is PEM. */ + int assume_base64; /* Assume input is base64 encoded. */ + + int identified; + int is_pem; + int is_base64; + int stop_seen; + int might_be_smime; + + int eof_seen; + + struct { + int idx; + unsigned char val; + int stop_seen; + } base64; +}; + + +/* Data used by the writer callbacks. */ +struct writer_cb_parm_s +{ + estream_t stream; /* Output stream. */ + + const char *pem_name; + + int wrote_begin; + int did_finish; + + struct { + int idx; + int quad_count; + unsigned char radbuf[4]; + } base64; + +}; + + +/* context for this module's functions */ +struct base64_context_s { + union { + struct reader_cb_parm_s rparm; + struct writer_cb_parm_s wparm; + } u; + + union { + ksba_reader_t reader; + ksba_writer_t writer; + } u2; +}; + + +/* The base-64 character list */ +static char bintoasc[64] = + "ABCDEFGHIJKLMNOPQRSTUVWXYZ" + "abcdefghijklmnopqrstuvwxyz" + "0123456789+/"; +/* The reverse base-64 list */ +static unsigned char asctobin[256] = { + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0x3e, 0xff, 0xff, 0xff, 0x3f, + 0x34, 0x35, 0x36, 0x37, 0x38, 0x39, 0x3a, 0x3b, 0x3c, 0x3d, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, + 0x07, 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, 0x0e, 0x0f, 0x10, 0x11, 0x12, + 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0x1a, 0x1b, 0x1c, 0x1d, 0x1e, 0x1f, 0x20, 0x21, 0x22, 0x23, 0x24, + 0x25, 0x26, 0x27, 0x28, 0x29, 0x2a, 0x2b, 0x2c, 0x2d, 0x2e, 0x2f, 0x30, + 0x31, 0x32, 0x33, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, + 0xff, 0xff, 0xff, 0xff +}; + + +static int +has_only_base64 (const unsigned char *line, int linelen) +{ + if (linelen < 20) + return 0; + for (; linelen; line++, linelen--) + { + if (*line == '\n' || (linelen > 1 && *line == '\r' && line[1] == '\n')) + break; + if ( !strchr (bintoasc, *line) ) + return 0; + } + return 1; /* yes */ +} + +static int +is_empty_line (const unsigned char *line, int linelen) +{ + if (linelen >= 2 && *line == '\r' && line[1] == '\n') + return 1; + if (linelen >= 1 && *line == '\n') + return 1; + return 0; +} + + +static int +base64_reader_cb (void *cb_value, char *buffer, size_t count, size_t *nread) +{ + struct reader_cb_parm_s *parm = cb_value; + size_t n; + int c, c2; + + *nread = 0; + if (!buffer) + return -1; /* not supported */ + + next: + if (!parm->linelen) + { + /* read an entire line or up to the size of the buffer */ + parm->line_counter++; + parm->have_lf = 0; + for (n=0; n < DIM(parm->line);) + { + c = es_getc (parm->fp); + if (c == EOF) + { + parm->eof_seen = 1; + if (es_ferror (parm->fp)) + return -1; + break; + } + parm->line[n++] = c; + if (c == '\n') + { + parm->have_lf = 1; + /* Fixme: we need to skip overlong lines while detecting + the dashed lines */ + break; + } + } + parm->linelen = n; + if (!n) + return -1; /* eof */ + parm->readpos = 0; + } + + if (!parm->identified) + { + if (!parm->autodetect) + { + if (parm->assume_pem) + { + /* wait for the header line */ + parm->linelen = parm->readpos = 0; + if (!parm->have_lf + || strncmp ((char*)parm->line, "-----BEGIN ", 11) + || !strncmp ((char*)parm->line+11, "PGP ", 4)) + goto next; + parm->is_pem = 1; + } + else if (parm->assume_base64) + parm->is_base64 = 1; + } + else if (parm->line_counter == 1 && !parm->have_lf) + { + /* first line too long - assume DER encoding */ + parm->is_pem = 0; + } + else if (parm->line_counter == 1 && parm->linelen && *parm->line == 0x30) + { + /* the very first byte does pretty much look like a SEQUENCE tag*/ + parm->is_pem = 0; + } + else if ( parm->have_lf + && !strncmp ((char*)parm->line, "-----BEGIN ", 11) + && strncmp ((char *)parm->line+11, "PGP ", 4) ) + { + /* Fixme: we must only compare if the line really starts at + the beginning */ + parm->is_pem = 1; + parm->linelen = parm->readpos = 0; + } + else if ( parm->have_lf && parm->line_counter == 1 + && parm->linelen >= 13 + && !ascii_memcasecmp (parm->line, "Content-Type:", 13)) + { /* might be a S/MIME body */ + parm->might_be_smime = 1; + parm->linelen = parm->readpos = 0; + goto next; + } + else if (parm->might_be_smime == 1 + && is_empty_line (parm->line, parm->linelen)) + { + parm->might_be_smime = 2; + parm->linelen = parm->readpos = 0; + goto next; + } + else if (parm->might_be_smime == 2) + { + parm->might_be_smime = 0; + if ( !has_only_base64 (parm->line, parm->linelen)) + { + parm->linelen = parm->readpos = 0; + goto next; + } + parm->is_pem = 1; + } + else + { + parm->linelen = parm->readpos = 0; + goto next; + } + parm->identified = 1; + parm->base64.stop_seen = 0; + parm->base64.idx = 0; + } + + + n = 0; + if (parm->is_pem || parm->is_base64) + { + if (parm->is_pem && parm->have_lf + && !strncmp ((char*)parm->line, "-----END ", 9)) + { + parm->identified = 0; + parm->linelen = parm->readpos = 0; + + /* If the caller want to read multiple PEM objects from one + file, we have to reset our internal state and return a + EOF immediately. The caller is the expected to use + ksba_reader_clear to clear the EOF condition and continue + to read. If we don't want to do that we just return 0 + bytes which will force the ksba_reader to skip until + EOF. */ + if (parm->allow_multi_pem) + { + parm->identified = 0; + parm->autodetect = 0; + parm->assume_pem = 1; + parm->stop_seen = 0; + return -1; /* Send EOF now. */ + } + } + else if (parm->stop_seen) + { /* skip the rest of the line */ + parm->linelen = parm->readpos = 0; + } + else + { + int idx = parm->base64.idx; + unsigned char val = parm->base64.val; + + while (n < count && parm->readpos < parm->linelen ) + { + c = parm->line[parm->readpos++]; + if (c == '\n' || c == ' ' || c == '\r' || c == '\t') + continue; + if (c == '=') + { /* pad character: stop */ + if (idx == 1) + buffer[n++] = val; + parm->stop_seen = 1; + break; + } + if( (c = asctobin[(c2=c)]) == 255 ) + { + log_error (_("invalid radix64 character %02x skipped\n"), + c2); + continue; + } + switch (idx) + { + case 0: + val = c << 2; + break; + case 1: + val |= (c>>4)&3; + buffer[n++] = val; + val = (c<<4)&0xf0; + break; + case 2: + val |= (c>>2)&15; + buffer[n++] = val; + val = (c<<6)&0xc0; + break; + case 3: + val |= c&0x3f; + buffer[n++] = val; + break; + } + idx = (idx+1) % 4; + } + if (parm->readpos == parm->linelen) + parm->linelen = parm->readpos = 0; + + parm->base64.idx = idx; + parm->base64.val = val; + } + } + else + { /* DER encoded */ + while (n < count && parm->readpos < parm->linelen) + buffer[n++] = parm->line[parm->readpos++]; + if (parm->readpos == parm->linelen) + parm->linelen = parm->readpos = 0; + } + + *nread = n; + return 0; +} + + + +static int +simple_reader_cb (void *cb_value, char *buffer, size_t count, size_t *nread) +{ + struct reader_cb_parm_s *parm = cb_value; + size_t n; + int c = 0; + + *nread = 0; + if (!buffer) + return -1; /* not supported */ + + for (n=0; n < count; n++) + { + c = es_getc (parm->fp); + if (c == EOF) + { + parm->eof_seen = 1; + if (es_ferror (parm->fp)) + return -1; + if (n) + break; /* Return what we have before an EOF. */ + return -1; + } + *(byte *)buffer++ = c; + } + + *nread = n; + return 0; +} + + + + +static int +base64_writer_cb (void *cb_value, const void *buffer, size_t count) +{ + struct writer_cb_parm_s *parm = cb_value; + unsigned char radbuf[4]; + int i, c, idx, quad_count; + const unsigned char *p; + estream_t stream = parm->stream; + + if (!count) + return 0; + + if (!parm->wrote_begin) + { + if (parm->pem_name) + { + es_fputs ("-----BEGIN ", stream); + es_fputs (parm->pem_name, stream); + es_fputs ("-----\n", stream); + } + parm->wrote_begin = 1; + parm->base64.idx = 0; + parm->base64.quad_count = 0; + } + + idx = parm->base64.idx; + quad_count = parm->base64.quad_count; + for (i=0; i < idx; i++) + radbuf[i] = parm->base64.radbuf[i]; + + for (p=buffer; count; p++, count--) + { + radbuf[idx++] = *p; + if (idx > 2) + { + idx = 0; + c = bintoasc[(*radbuf >> 2) & 077]; + es_putc (c, stream); + c = bintoasc[(((*radbuf<<4)&060)|((radbuf[1] >> 4)&017))&077]; + es_putc (c, stream); + c = bintoasc[(((radbuf[1]<<2)&074)|((radbuf[2]>>6)&03))&077]; + es_putc (c, stream); + c = bintoasc[radbuf[2]&077]; + es_putc (c, stream); + if (++quad_count >= (64/4)) + { + es_fputs (LF, stream); + quad_count = 0; + } + } + } + for (i=0; i < idx; i++) + parm->base64.radbuf[i] = radbuf[i]; + parm->base64.idx = idx; + parm->base64.quad_count = quad_count; + + return es_ferror (stream)? gpg_error_from_syserror () : 0; +} + + +/* This callback is only used in stream mode. Hiowever, we don't + restrict it to this. */ +static int +plain_writer_cb (void *cb_value, const void *buffer, size_t count) +{ + struct writer_cb_parm_s *parm = cb_value; + estream_t stream = parm->stream; + + if (!count) + return 0; + + es_write (stream, buffer, count, NULL); + + return es_ferror (stream)? gpg_error_from_syserror () : 0; +} + + +static int +base64_finish_write (struct writer_cb_parm_s *parm) +{ + unsigned char *radbuf; + int c, idx, quad_count; + estream_t stream = parm->stream; + + if (!parm->wrote_begin) + return 0; /* Nothing written or we are not called in base-64 mode. */ + + /* flush the base64 encoding */ + idx = parm->base64.idx; + quad_count = parm->base64.quad_count; + if (idx) + { + radbuf = parm->base64.radbuf; + + c = bintoasc[(*radbuf>>2)&077]; + es_putc (c, stream); + if (idx == 1) + { + c = bintoasc[((*radbuf << 4) & 060) & 077]; + es_putc (c, stream); + es_putc ('=', stream); + es_putc ('=', stream); + } + else + { + c = bintoasc[(((*radbuf<<4)&060)|((radbuf[1]>>4)&017))&077]; + es_putc (c, stream); + c = bintoasc[((radbuf[1] << 2) & 074) & 077]; + es_putc (c, stream); + es_putc ('=', stream); + + } + if (++quad_count >= (64/4)) + { + es_fputs (LF, stream); + quad_count = 0; + } + } + + if (quad_count) + es_fputs (LF, stream); + + if (parm->pem_name) + { + es_fputs ("-----END ", stream); + es_fputs (parm->pem_name, stream); + es_fputs ("-----\n", stream); + } + + return es_ferror (stream)? gpg_error_from_syserror () : 0; +} + + + + +/* Create a reader for the given file descriptor. Depending on the + control information an input decoding is automagically chosen. + The function returns a Base64Context object which must be passed to + the gpgme_destroy_reader function. The created KsbaReader object + is also returned, but the caller must not call the + ksba_reader_release function on. If ALLOW_MULTI_PEM is true, the + reader expects that the caller uses ksba_reader_clear after EOF + until no more objects were found. */ +int +gpgsm_create_reader (Base64Context *ctx, + ctrl_t ctrl, estream_t fp, int allow_multi_pem, + ksba_reader_t *r_reader) +{ + int rc; + ksba_reader_t r; + + *r_reader = NULL; + *ctx = xtrycalloc (1, sizeof **ctx); + if (!*ctx) + return out_of_core (); + (*ctx)->u.rparm.allow_multi_pem = allow_multi_pem; + + rc = ksba_reader_new (&r); + if (rc) + { + xfree (*ctx); *ctx = NULL; + return rc; + } + + (*ctx)->u.rparm.fp = fp; + if (ctrl->is_pem) + { + (*ctx)->u.rparm.assume_pem = 1; + (*ctx)->u.rparm.assume_base64 = 1; + rc = ksba_reader_set_cb (r, base64_reader_cb, &(*ctx)->u.rparm); + } + else if (ctrl->is_base64) + { + (*ctx)->u.rparm.assume_base64 = 1; + rc = ksba_reader_set_cb (r, base64_reader_cb, &(*ctx)->u.rparm); + } + else if (ctrl->autodetect_encoding) + { + (*ctx)->u.rparm.autodetect = 1; + rc = ksba_reader_set_cb (r, base64_reader_cb, &(*ctx)->u.rparm); + } + else + rc = ksba_reader_set_cb (r, simple_reader_cb, &(*ctx)->u.rparm); + + if (rc) + { + ksba_reader_release (r); + xfree (*ctx); *ctx = NULL; + return rc; + } + + (*ctx)->u2.reader = r; + *r_reader = r; + return 0; +} + + +int +gpgsm_reader_eof_seen (Base64Context ctx) +{ + return ctx && ctx->u.rparm.eof_seen; +} + +void +gpgsm_destroy_reader (Base64Context ctx) +{ + if (!ctx) + return; + + ksba_reader_release (ctx->u2.reader); + xfree (ctx); +} + + + +/* Create a writer for the given STREAM. Depending on + the control information an output encoding is automagically + chosen. The function returns a Base64Context object which must be + passed to the gpgme_destroy_writer function. The created + KsbaWriter object is also returned, but the caller must not call + the ksba_reader_release function on it. */ +int +gpgsm_create_writer (Base64Context *ctx, ctrl_t ctrl, estream_t stream, + ksba_writer_t *r_writer) +{ + int rc; + ksba_writer_t w; + + *r_writer = NULL; + *ctx = xtrycalloc (1, sizeof **ctx); + if (!*ctx) + return out_of_core (); + + rc = ksba_writer_new (&w); + if (rc) + { + xfree (*ctx); *ctx = NULL; + return rc; + } + + if (ctrl->create_pem || ctrl->create_base64) + { + (*ctx)->u.wparm.stream = stream; + if (ctrl->create_pem) + (*ctx)->u.wparm.pem_name = ctrl->pem_name? ctrl->pem_name + : "CMS OBJECT"; + rc = ksba_writer_set_cb (w, base64_writer_cb, &(*ctx)->u.wparm); + } + else if (stream) + { + (*ctx)->u.wparm.stream = stream; + rc = ksba_writer_set_cb (w, plain_writer_cb, &(*ctx)->u.wparm); + } + else + rc = gpg_error (GPG_ERR_INV_ARG); + + if (rc) + { + ksba_writer_release (w); + xfree (*ctx); *ctx = NULL; + return rc; + } + + (*ctx)->u2.writer = w; + *r_writer = w; + return 0; +} + + +int +gpgsm_finish_writer (Base64Context ctx) +{ + struct writer_cb_parm_s *parm; + + if (!ctx) + return gpg_error (GPG_ERR_INV_VALUE); + parm = &ctx->u.wparm; + if (parm->did_finish) + return 0; /* Already done. */ + parm->did_finish = 1; + if (!parm->stream) + return 0; /* Callback was not used. */ + return base64_finish_write (parm); +} + +void +gpgsm_destroy_writer (Base64Context ctx) +{ + if (!ctx) + return; + + ksba_writer_release (ctx->u2.writer); + xfree (ctx); +} |