security-and-privacy/gnupg
security-and-privacy/gnupg
1
0
Fork 0
mirror of git://git.gnupg.org/gnupg.git synced 2025-07-02 22:46:30 +02:00
Code Activity

gpg,gpgsm: Block signals during keyring/keybox update.

Browse source 
* kbx/keybox-util.c (keybox_file_rename): Add arg BLOCK_SIGNALS.
* kbx/keybox-update.c (rename_tmp_file): Block all signals when doing
a double rename.
* g10/keyring.c (rename_tmp_file): Block all signals during the double
rename.
--

This might fix
Debian-bug-id: 831510

Signed-off-by: Werner Koch <wk@gnupg.org>
This commit is contained in:
Werner Koch 2016-08-03 15:31:27 +02:00
parent 3a2421c940
commit 48a2c93a18
No known key found for this signature in database
GPG key ID: E3FDFF218E45B72B
4 changed files with 88 additions and 53 deletions
Download patch file Download diff file Expand all files Collapse all files

88
kbx/keybox-util.c
View file

@ -27,6 +27,7 @@
#endif
#include "keybox-defs.h"
#include "utilproto.h"
static void *(*alloc_func)(size_t n) = malloc;
@ -147,55 +148,70 @@ keybox_tmp_names (const char *filename, int for_keyring,
}
/* Wrapper for rename(2) to handle Windows peculiarities. */
/* Wrapper for rename(2) to handle Windows peculiarities. If
* BLOCK_SIGNALS is not NULL and points to a variable set to true, all
* signals will be blocked by calling gnupg_block_all_signals; the
* caller needs to call gnupg_unblock_all_signals if that variable is
* still set to true on return. */
gpg_error_t
keybox_file_rename (const char *oldname, const char *newname)
keybox_file_rename (const char *oldname, const char *newname,
int *block_signals)
{
gpg_error_t err = 0;
if (block_signals && *block_signals)
gnupg_block_all_signals ();
#ifdef HAVE_DOSISH_SYSTEM
int wtime = 0;
{
int wtime = 0;
gnupg_remove (newname);
again:
if (rename (oldname, newname))
{
if (GetLastError () == ERROR_SHARING_VIOLATION)
{
/* Another process has the file open. We do not use a lock
* for read but instead we wait until the other process has
* closed the file. This may take long but that would also
* be the case with a dotlock approach for read and write.
* Note that we don't need this on Unix due to the inode
* concept.
*
* So let's wait until the rename has worked. The retry
* intervals are 50, 100, 200, 400, 800, 50ms, ... */
if (!wtime || wtime >= 800)
wtime = 50;
else
wtime *= 2;
gnupg_remove (newname);
again:
if (rename (oldname, newname))
{
if (GetLastError () == ERROR_SHARING_VIOLATION)
{
/* Another process has the file open. We do not use a
* lock for read but instead we wait until the other
* process has closed the file. This may take long but
* that would also be the case with a dotlock approach for
* read and write. Note that we don't need this on Unix
* due to the inode concept.
*
* So let's wait until the rename has worked. The retry
* intervals are 50, 100, 200, 400, 800, 50ms, ... */
if (!wtime || wtime >= 800)
wtime = 50;
else
wtime *= 2;
if (wtime >= 800)
log_info ("waiting for file '%s' to become accessible ...\n",
oldname);
Sleep (wtime);
goto again;
}
err = gpg_error_from_syserror ();
}
if (wtime >= 800)
log_info ("waiting for file '%s' to become accessible ...\n",
oldname);
Sleep (wtime);
goto again;
}
err = gpg_error_from_syserror ();
}
}
#else /* Unix */
{
#ifdef __riscos__
gnupg_remove (newname);
gnupg_remove (newname);
#endif
if (rename (oldname, newname) )
err = gpg_error_from_syserror ();
if (rename (oldname, newname) )
err = gpg_error_from_syserror ();
}
#endif /* Unix */
if (block_signals && *block_signals && err)
{
gnupg_unblock_all_signals ();
*block_signals = 0;
}
if (err)
log_error ("renaming '%s' to '%s' failed: %s\n",
oldname, newname, gpg_strerror (err));