1
0
Fork 0
mirror of https://github.com/git/git.git synced 2024-05-13 07:36:09 +02:00
git/tempfile.c

319 lines
8.3 KiB
C
Raw Normal View History

/*
* State diagram and cleanup
* -------------------------
*
* If the program exits while a temporary file is active, we want to
* make sure that we remove it. This is done by remembering the active
* temporary files in a linked list, `tempfile_list`. An `atexit(3)`
* handler and a signal handler are registered, to clean up any active
* temporary files.
*
* Because the signal handler can run at any time, `tempfile_list` and
* the `tempfile` objects that comprise it must be kept in
* self-consistent states at all times.
*
* The possible states of a `tempfile` object are as follows:
*
* - Uninitialized. In this state the object's `on_list` field must be
* zero but the rest of its contents need not be initialized. As
* soon as the object is used in any way, it is irrevocably
* registered in `tempfile_list`, and `on_list` is set.
*
* - Active, file open (after `create_tempfile()` or
* `reopen_tempfile()`). In this state:
*
* - the temporary file exists
* - `active` is set
* - `filename` holds the filename of the temporary file
* - `fd` holds a file descriptor open for writing to it
* - `fp` holds a pointer to an open `FILE` object if and only if
* `fdopen_tempfile()` has been called on the object
* - `owner` holds the PID of the process that created the file
*
tempfile: do not delete tempfile on failed close When close_tempfile() fails, we delete the tempfile and reset the fields of the tempfile struct. This makes it easier for callers to return without cleaning up, but it also makes this common pattern: if (close_tempfile(tempfile)) return error_errno("error closing %s", tempfile->filename.buf); wrong, because the "filename" field has been reset after the failed close. And it's not easy to fix, as in many cases we don't have another copy of the filename (e.g., if it was created via one of the mks_tempfile functions, and we just have the original template string). Let's drop the feature that a failed close automatically deletes the file. This puts the burden on the caller to do the deletion themselves, but this isn't that big a deal. Callers which do: if (write(...) || close_tempfile(...)) { delete_tempfile(...); return -1; } already had to call delete when the write() failed, and so aren't affected. Likewise, any caller which just calls die() in the error path is OK; we'll delete the tempfile during the atexit handler. Because this patch changes the semantics of close_tempfile() without changing its signature, all callers need to be manually checked and converted to the new scheme. This patch covers all in-tree callers, but there may be others for not-yet-merged topics. To catch these, we rename the function to close_tempfile_gently(), which will attract compile-time attention to new callers. (Technically the original could be considered "gentle" already in that it didn't die() on errors, but this one is even more so). Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:30 +02:00
* - Active, file closed (after `close_tempfile_gently()`). Same
* as the previous state, except that the temporary file is closed,
* `fd` is -1, and `fp` is `NULL`.
*
tempfile: do not delete tempfile on failed close When close_tempfile() fails, we delete the tempfile and reset the fields of the tempfile struct. This makes it easier for callers to return without cleaning up, but it also makes this common pattern: if (close_tempfile(tempfile)) return error_errno("error closing %s", tempfile->filename.buf); wrong, because the "filename" field has been reset after the failed close. And it's not easy to fix, as in many cases we don't have another copy of the filename (e.g., if it was created via one of the mks_tempfile functions, and we just have the original template string). Let's drop the feature that a failed close automatically deletes the file. This puts the burden on the caller to do the deletion themselves, but this isn't that big a deal. Callers which do: if (write(...) || close_tempfile(...)) { delete_tempfile(...); return -1; } already had to call delete when the write() failed, and so aren't affected. Likewise, any caller which just calls die() in the error path is OK; we'll delete the tempfile during the atexit handler. Because this patch changes the semantics of close_tempfile() without changing its signature, all callers need to be manually checked and converted to the new scheme. This patch covers all in-tree callers, but there may be others for not-yet-merged topics. To catch these, we rename the function to close_tempfile_gently(), which will attract compile-time attention to new callers. (Technically the original could be considered "gentle" already in that it didn't die() on errors, but this one is even more so). Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:30 +02:00
* - Inactive (after `delete_tempfile()`, `rename_tempfile()`, or a
* failed attempt to create a temporary file). In this state:
*
* - `active` is unset
* - `filename` is empty (usually, though there are transitory
* states in which this condition doesn't hold). Client code should
* *not* rely on the filename being empty in this state.
* - `fd` is -1 and `fp` is `NULL`
* - the object is left registered in the `tempfile_list`, and
* `on_list` is set.
*
* A temporary file is owned by the process that created it. The
* `tempfile` has an `owner` field that records the owner's PID. This
* field is used to prevent a forked process from deleting a temporary
* file created by its parent.
*/
#include "cache.h"
#include "tempfile.h"
#include "sigchain.h"
static struct tempfile *volatile tempfile_list;
tempfile: robustify cleanup handler We may call remove_tempfiles() from an atexit handler, or from a signal handler. In the latter case we must take care to avoid functions which may deadlock if the process is in an unknown state, including looking at any stdio handles (which may be in the middle of doing I/O and locked) or calling malloc() or free(). The current implementation calls delete_tempfile(). We unset the tempfile's stdio handle (if any) to avoid deadlocking there. But delete_tempfile() still calls unlink_or_warn(), which can deadlock writing to stderr if the unlink fails. Since delete_tempfile() isn't very long, let's just open-code our own simple conservative version of the same thing. Notably: 1. The "skip_fclose" flag is now called "in_signal_handler", because it should inform more decisions than just the fclose handling. 2. We can replace close_tempfile() with just close(fd). That skips the fclose() question altogether. This is fine for the atexit() case, too; there's no point flushing data to a file which we're about to delete anyway. 3. We can choose between unlink/unlink_or_warn based on whether it's safe to use stderr. 4. We can replace the deactivate_tempfile() call with a simple setting of the active flag. There's no need to do any further cleanup since we know the program is exiting. And even though the current deactivation code is safe in a signal handler, this frees us up in future patches to make non-signal deactivation more complicated (e.g., by freeing resources). 5. There's no need to remove items from the tempfile_list. The "active" flag is the ultimate answer to whether an entry has been handled or not. Manipulating the list just introduces more chance of recursive signals stomping on each other, and the whole list will go away when the program exits anyway. Less is more. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:53 +02:00
static void remove_tempfiles(int in_signal_handler)
{
pid_t me = getpid();
tempfile: robustify cleanup handler We may call remove_tempfiles() from an atexit handler, or from a signal handler. In the latter case we must take care to avoid functions which may deadlock if the process is in an unknown state, including looking at any stdio handles (which may be in the middle of doing I/O and locked) or calling malloc() or free(). The current implementation calls delete_tempfile(). We unset the tempfile's stdio handle (if any) to avoid deadlocking there. But delete_tempfile() still calls unlink_or_warn(), which can deadlock writing to stderr if the unlink fails. Since delete_tempfile() isn't very long, let's just open-code our own simple conservative version of the same thing. Notably: 1. The "skip_fclose" flag is now called "in_signal_handler", because it should inform more decisions than just the fclose handling. 2. We can replace close_tempfile() with just close(fd). That skips the fclose() question altogether. This is fine for the atexit() case, too; there's no point flushing data to a file which we're about to delete anyway. 3. We can choose between unlink/unlink_or_warn based on whether it's safe to use stderr. 4. We can replace the deactivate_tempfile() call with a simple setting of the active flag. There's no need to do any further cleanup since we know the program is exiting. And even though the current deactivation code is safe in a signal handler, this frees us up in future patches to make non-signal deactivation more complicated (e.g., by freeing resources). 5. There's no need to remove items from the tempfile_list. The "active" flag is the ultimate answer to whether an entry has been handled or not. Manipulating the list just introduces more chance of recursive signals stomping on each other, and the whole list will go away when the program exits anyway. Less is more. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:53 +02:00
struct tempfile *volatile p;
tempfile: robustify cleanup handler We may call remove_tempfiles() from an atexit handler, or from a signal handler. In the latter case we must take care to avoid functions which may deadlock if the process is in an unknown state, including looking at any stdio handles (which may be in the middle of doing I/O and locked) or calling malloc() or free(). The current implementation calls delete_tempfile(). We unset the tempfile's stdio handle (if any) to avoid deadlocking there. But delete_tempfile() still calls unlink_or_warn(), which can deadlock writing to stderr if the unlink fails. Since delete_tempfile() isn't very long, let's just open-code our own simple conservative version of the same thing. Notably: 1. The "skip_fclose" flag is now called "in_signal_handler", because it should inform more decisions than just the fclose handling. 2. We can replace close_tempfile() with just close(fd). That skips the fclose() question altogether. This is fine for the atexit() case, too; there's no point flushing data to a file which we're about to delete anyway. 3. We can choose between unlink/unlink_or_warn based on whether it's safe to use stderr. 4. We can replace the deactivate_tempfile() call with a simple setting of the active flag. There's no need to do any further cleanup since we know the program is exiting. And even though the current deactivation code is safe in a signal handler, this frees us up in future patches to make non-signal deactivation more complicated (e.g., by freeing resources). 5. There's no need to remove items from the tempfile_list. The "active" flag is the ultimate answer to whether an entry has been handled or not. Manipulating the list just introduces more chance of recursive signals stomping on each other, and the whole list will go away when the program exits anyway. Less is more. Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:53 +02:00
for (p = tempfile_list; p; p = p->next) {
if (!is_tempfile_active(p) || p->owner != me)
continue;
if (p->fd >= 0)
close(p->fd);
if (in_signal_handler)
unlink(p->filename.buf);
else
unlink_or_warn(p->filename.buf);
p->active = 0;
}
}
static void remove_tempfiles_on_exit(void)
{
remove_tempfiles(0);
}
static void remove_tempfiles_on_signal(int signo)
{
remove_tempfiles(1);
sigchain_pop(signo);
raise(signo);
}
/*
* Initialize *tempfile if necessary and add it to tempfile_list.
*/
static void prepare_tempfile_object(struct tempfile *tempfile)
{
if (!tempfile_list) {
/* One-time initialization */
sigchain_push_common(remove_tempfiles_on_signal);
atexit(remove_tempfiles_on_exit);
}
if (is_tempfile_active(tempfile))
BUG("prepare_tempfile_object called for active object");
if (!tempfile->on_list) {
/* Initialize *tempfile and add it to tempfile_list: */
tempfile->fd = -1;
tempfile->fp = NULL;
tempfile->active = 0;
tempfile->owner = 0;
strbuf_init(&tempfile->filename, 0);
tempfile->next = tempfile_list;
tempfile_list = tempfile;
tempfile->on_list = 1;
} else if (tempfile->filename.len) {
/* This shouldn't happen, but better safe than sorry. */
BUG("prepare_tempfile_object called for improperly-reset object");
}
}
static void activate_tempfile(struct tempfile *tempfile)
{
tempfile->owner = getpid();
tempfile->active = 1;
}
static void deactivate_tempfile(struct tempfile *tempfile)
{
tempfile->active = 0;
strbuf_reset(&tempfile->filename);
}
/* Make sure errno contains a meaningful value on error */
int create_tempfile(struct tempfile *tempfile, const char *path)
{
prepare_tempfile_object(tempfile);
strbuf_add_absolute_path(&tempfile->filename, path);
2016-08-22 14:47:55 +02:00
tempfile->fd = open(tempfile->filename.buf,
O_RDWR | O_CREAT | O_EXCL | O_CLOEXEC, 0666);
if (O_CLOEXEC && tempfile->fd < 0 && errno == EINVAL)
/* Try again w/o O_CLOEXEC: the kernel might not support it */
tempfile->fd = open(tempfile->filename.buf,
O_RDWR | O_CREAT | O_EXCL, 0666);
if (tempfile->fd < 0) {
deactivate_tempfile(tempfile);
return -1;
}
activate_tempfile(tempfile);
if (adjust_shared_perm(tempfile->filename.buf)) {
int save_errno = errno;
error("cannot fix permission bits on %s", tempfile->filename.buf);
delete_tempfile(tempfile);
errno = save_errno;
return -1;
}
return tempfile->fd;
}
void register_tempfile(struct tempfile *tempfile, const char *path)
{
prepare_tempfile_object(tempfile);
strbuf_add_absolute_path(&tempfile->filename, path);
activate_tempfile(tempfile);
}
int mks_tempfile_sm(struct tempfile *tempfile,
const char *template, int suffixlen, int mode)
{
prepare_tempfile_object(tempfile);
strbuf_add_absolute_path(&tempfile->filename, template);
tempfile->fd = git_mkstemps_mode(tempfile->filename.buf, suffixlen, mode);
if (tempfile->fd < 0) {
deactivate_tempfile(tempfile);
return -1;
}
activate_tempfile(tempfile);
return tempfile->fd;
}
int mks_tempfile_tsm(struct tempfile *tempfile,
const char *template, int suffixlen, int mode)
{
const char *tmpdir;
prepare_tempfile_object(tempfile);
tmpdir = getenv("TMPDIR");
if (!tmpdir)
tmpdir = "/tmp";
strbuf_addf(&tempfile->filename, "%s/%s", tmpdir, template);
tempfile->fd = git_mkstemps_mode(tempfile->filename.buf, suffixlen, mode);
if (tempfile->fd < 0) {
deactivate_tempfile(tempfile);
return -1;
}
activate_tempfile(tempfile);
return tempfile->fd;
}
int xmks_tempfile_m(struct tempfile *tempfile, const char *template, int mode)
{
int fd;
struct strbuf full_template = STRBUF_INIT;
strbuf_add_absolute_path(&full_template, template);
fd = mks_tempfile_m(tempfile, full_template.buf, mode);
if (fd < 0)
die_errno("Unable to create temporary file '%s'",
full_template.buf);
strbuf_release(&full_template);
return fd;
}
FILE *fdopen_tempfile(struct tempfile *tempfile, const char *mode)
{
if (!is_tempfile_active(tempfile))
BUG("fdopen_tempfile() called for inactive object");
if (tempfile->fp)
BUG("fdopen_tempfile() called for open object");
tempfile->fp = fdopen(tempfile->fd, mode);
return tempfile->fp;
}
const char *get_tempfile_path(struct tempfile *tempfile)
{
if (!is_tempfile_active(tempfile))
BUG("get_tempfile_path() called for inactive object");
return tempfile->filename.buf;
}
int get_tempfile_fd(struct tempfile *tempfile)
{
if (!is_tempfile_active(tempfile))
BUG("get_tempfile_fd() called for inactive object");
return tempfile->fd;
}
FILE *get_tempfile_fp(struct tempfile *tempfile)
{
if (!is_tempfile_active(tempfile))
BUG("get_tempfile_fp() called for inactive object");
return tempfile->fp;
}
tempfile: do not delete tempfile on failed close When close_tempfile() fails, we delete the tempfile and reset the fields of the tempfile struct. This makes it easier for callers to return without cleaning up, but it also makes this common pattern: if (close_tempfile(tempfile)) return error_errno("error closing %s", tempfile->filename.buf); wrong, because the "filename" field has been reset after the failed close. And it's not easy to fix, as in many cases we don't have another copy of the filename (e.g., if it was created via one of the mks_tempfile functions, and we just have the original template string). Let's drop the feature that a failed close automatically deletes the file. This puts the burden on the caller to do the deletion themselves, but this isn't that big a deal. Callers which do: if (write(...) || close_tempfile(...)) { delete_tempfile(...); return -1; } already had to call delete when the write() failed, and so aren't affected. Likewise, any caller which just calls die() in the error path is OK; we'll delete the tempfile during the atexit handler. Because this patch changes the semantics of close_tempfile() without changing its signature, all callers need to be manually checked and converted to the new scheme. This patch covers all in-tree callers, but there may be others for not-yet-merged topics. To catch these, we rename the function to close_tempfile_gently(), which will attract compile-time attention to new callers. (Technically the original could be considered "gentle" already in that it didn't die() on errors, but this one is even more so). Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:30 +02:00
int close_tempfile_gently(struct tempfile *tempfile)
{
int fd;
FILE *fp;
int err;
if (!is_tempfile_active(tempfile) || tempfile->fd < 0)
return 0;
fd = tempfile->fd;
fp = tempfile->fp;
tempfile->fd = -1;
if (fp) {
tempfile->fp = NULL;
if (ferror(fp)) {
err = -1;
if (!fclose(fp))
errno = EIO;
} else {
err = fclose(fp);
}
} else {
err = close(fd);
}
tempfile: do not delete tempfile on failed close When close_tempfile() fails, we delete the tempfile and reset the fields of the tempfile struct. This makes it easier for callers to return without cleaning up, but it also makes this common pattern: if (close_tempfile(tempfile)) return error_errno("error closing %s", tempfile->filename.buf); wrong, because the "filename" field has been reset after the failed close. And it's not easy to fix, as in many cases we don't have another copy of the filename (e.g., if it was created via one of the mks_tempfile functions, and we just have the original template string). Let's drop the feature that a failed close automatically deletes the file. This puts the burden on the caller to do the deletion themselves, but this isn't that big a deal. Callers which do: if (write(...) || close_tempfile(...)) { delete_tempfile(...); return -1; } already had to call delete when the write() failed, and so aren't affected. Likewise, any caller which just calls die() in the error path is OK; we'll delete the tempfile during the atexit handler. Because this patch changes the semantics of close_tempfile() without changing its signature, all callers need to be manually checked and converted to the new scheme. This patch covers all in-tree callers, but there may be others for not-yet-merged topics. To catch these, we rename the function to close_tempfile_gently(), which will attract compile-time attention to new callers. (Technically the original could be considered "gentle" already in that it didn't die() on errors, but this one is even more so). Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:30 +02:00
return err ? -1 : 0;
}
int reopen_tempfile(struct tempfile *tempfile)
{
if (!is_tempfile_active(tempfile))
BUG("reopen_tempfile called for an inactive object");
if (0 <= tempfile->fd)
BUG("reopen_tempfile called for an open object");
tempfile->fd = open(tempfile->filename.buf, O_WRONLY);
return tempfile->fd;
}
int rename_tempfile(struct tempfile *tempfile, const char *path)
{
if (!is_tempfile_active(tempfile))
BUG("rename_tempfile called for inactive object");
tempfile: do not delete tempfile on failed close When close_tempfile() fails, we delete the tempfile and reset the fields of the tempfile struct. This makes it easier for callers to return without cleaning up, but it also makes this common pattern: if (close_tempfile(tempfile)) return error_errno("error closing %s", tempfile->filename.buf); wrong, because the "filename" field has been reset after the failed close. And it's not easy to fix, as in many cases we don't have another copy of the filename (e.g., if it was created via one of the mks_tempfile functions, and we just have the original template string). Let's drop the feature that a failed close automatically deletes the file. This puts the burden on the caller to do the deletion themselves, but this isn't that big a deal. Callers which do: if (write(...) || close_tempfile(...)) { delete_tempfile(...); return -1; } already had to call delete when the write() failed, and so aren't affected. Likewise, any caller which just calls die() in the error path is OK; we'll delete the tempfile during the atexit handler. Because this patch changes the semantics of close_tempfile() without changing its signature, all callers need to be manually checked and converted to the new scheme. This patch covers all in-tree callers, but there may be others for not-yet-merged topics. To catch these, we rename the function to close_tempfile_gently(), which will attract compile-time attention to new callers. (Technically the original could be considered "gentle" already in that it didn't die() on errors, but this one is even more so). Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:30 +02:00
if (close_tempfile_gently(tempfile)) {
delete_tempfile(tempfile);
return -1;
tempfile: do not delete tempfile on failed close When close_tempfile() fails, we delete the tempfile and reset the fields of the tempfile struct. This makes it easier for callers to return without cleaning up, but it also makes this common pattern: if (close_tempfile(tempfile)) return error_errno("error closing %s", tempfile->filename.buf); wrong, because the "filename" field has been reset after the failed close. And it's not easy to fix, as in many cases we don't have another copy of the filename (e.g., if it was created via one of the mks_tempfile functions, and we just have the original template string). Let's drop the feature that a failed close automatically deletes the file. This puts the burden on the caller to do the deletion themselves, but this isn't that big a deal. Callers which do: if (write(...) || close_tempfile(...)) { delete_tempfile(...); return -1; } already had to call delete when the write() failed, and so aren't affected. Likewise, any caller which just calls die() in the error path is OK; we'll delete the tempfile during the atexit handler. Because this patch changes the semantics of close_tempfile() without changing its signature, all callers need to be manually checked and converted to the new scheme. This patch covers all in-tree callers, but there may be others for not-yet-merged topics. To catch these, we rename the function to close_tempfile_gently(), which will attract compile-time attention to new callers. (Technically the original could be considered "gentle" already in that it didn't die() on errors, but this one is even more so). Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:30 +02:00
}
if (rename(tempfile->filename.buf, path)) {
int save_errno = errno;
delete_tempfile(tempfile);
errno = save_errno;
return -1;
}
deactivate_tempfile(tempfile);
return 0;
}
void delete_tempfile(struct tempfile *tempfile)
{
if (!is_tempfile_active(tempfile))
return;
tempfile: do not delete tempfile on failed close When close_tempfile() fails, we delete the tempfile and reset the fields of the tempfile struct. This makes it easier for callers to return without cleaning up, but it also makes this common pattern: if (close_tempfile(tempfile)) return error_errno("error closing %s", tempfile->filename.buf); wrong, because the "filename" field has been reset after the failed close. And it's not easy to fix, as in many cases we don't have another copy of the filename (e.g., if it was created via one of the mks_tempfile functions, and we just have the original template string). Let's drop the feature that a failed close automatically deletes the file. This puts the burden on the caller to do the deletion themselves, but this isn't that big a deal. Callers which do: if (write(...) || close_tempfile(...)) { delete_tempfile(...); return -1; } already had to call delete when the write() failed, and so aren't affected. Likewise, any caller which just calls die() in the error path is OK; we'll delete the tempfile during the atexit handler. Because this patch changes the semantics of close_tempfile() without changing its signature, all callers need to be manually checked and converted to the new scheme. This patch covers all in-tree callers, but there may be others for not-yet-merged topics. To catch these, we rename the function to close_tempfile_gently(), which will attract compile-time attention to new callers. (Technically the original could be considered "gentle" already in that it didn't die() on errors, but this one is even more so). Signed-off-by: Jeff King <peff@peff.net> Signed-off-by: Junio C Hamano <gitster@pobox.com>
2017-09-05 14:14:30 +02:00
close_tempfile_gently(tempfile);
unlink_or_warn(tempfile->filename.buf);
deactivate_tempfile(tempfile);
}