Message ID | 1433541647-300-1-git-send-email-nicholson@endlessm.com (mailing list archive) |
---|---|
State | Rejected |
Delegated to: | Takashi Iwai |
Headers | show |
At Fri, 5 Jun 2015 15:00:47 -0700, Dan Nicholson wrote: > > Try to create the directory for the state file when saving so we don't > depend on it being created ahead of time. This only checks for failures > on existing directories and doesn't try to create the leading > directories or workaround any other errors. This should catch the common > case where /var/lib exists, but /var/lib/alsa doesn't. I don't think it's the role of alsactl. It saves a file on the certain directory. If it doesn't exist, it's a failure of the installed package. thanks, Takashi > > Signed-off-by: Dan Nicholson <nicholson@endlessm.com> > --- > alsactl/state.c | 23 +++++++++++++++++++++++ > 1 file changed, 23 insertions(+) > > diff --git a/alsactl/state.c b/alsactl/state.c > index 3908ec4..8ca3d6e 100644 > --- a/alsactl/state.c > +++ b/alsactl/state.c > @@ -27,6 +27,9 @@ > #include <stdio.h> > #include <assert.h> > #include <errno.h> > +#include <string.h> > +#include <sys/stat.h> > +#include <sys/types.h> > #include <alsa/asoundlib.h> > #include "alsactl.h" > > @@ -1544,6 +1547,7 @@ int save_state(const char *file, const char *cardname) > snd_output_t *out; > int stdio; > char *nfile = NULL; > + char *filedir = NULL; > int lock_fd = -EINVAL; > > err = snd_config_top(&config); > @@ -1553,6 +1557,9 @@ int save_state(const char *file, const char *cardname) > } > stdio = !strcmp(file, "-"); > if (!stdio) { > + char *tmp; > + mode_t mode = S_IRWXU | S_IRWXG | S_IRWXO; > + > nfile = malloc(strlen(file) + 5); > if (nfile == NULL) { > error("No enough memory..."); > @@ -1561,6 +1568,21 @@ int save_state(const char *file, const char *cardname) > } > strcpy(nfile, file); > strcat(nfile, ".new"); > + filedir = strdup(file); > + if (filedir == NULL) { > + error("Not enough memory..."); > + err = -ENOMEM; > + goto out; > + } > + tmp = strrchr(filedir, '/'); > + if (tmp && tmp != filedir) { > + *tmp = '\0'; > + if (mkdir(filedir, mode) != 0 && errno != EEXIST) { > + error("Could not create directory %s: %s", > + filedir, strerror(errno)); > + goto out; > + } > + } > lock_fd = state_lock(file, 10); > if (lock_fd < 0) { > err = lock_fd; > @@ -1640,6 +1662,7 @@ out: > if (!stdio && lock_fd >= 0) > state_unlock(lock_fd, file); > free(nfile); > + free(filedir); > snd_config_delete(config); > snd_config_update_free_global(); > return err; > -- > 1.8.1.2 > > _______________________________________________ > Alsa-devel mailing list > Alsa-devel@alsa-project.org > http://mailman.alsa-project.org/mailman/listinfo/alsa-devel >
On Jun 8, 2015 4:38 AM, "Takashi Iwai" <tiwai@suse.de> wrote: > > At Fri, 5 Jun 2015 15:00:47 -0700, > Dan Nicholson wrote: > > > > Try to create the directory for the state file when saving so we don't > > depend on it being created ahead of time. This only checks for failures > > on existing directories and doesn't try to create the leading > > directories or workaround any other errors. This should catch the common > > case where /var/lib exists, but /var/lib/alsa doesn't. > > I don't think it's the role of alsactl. It saves a file on the > certain directory. If it doesn't exist, it's a failure of the > installed package. Sure, that's understandable, but there's a couple reasons I think this is helpful addition. First, if no path is supplied, store will save to /var/lib/alsa. So, it's not as of the user has supplied a path it didn't setup correctly. It would be nice if alsactl worked out of the box without additional integration by packagers. Second, my real motivation for fixing this is to support stateless type of systems that come with a clean /var. At Endless we're using ostree. The OS is composed by Debian packages, and indeed alsa-utils is setup to create /var/lib/alsa on install. However, to use the same OS snapshot for all users, the contents of /var are stripped since they represent local system state. We can certainly add a method for creating the directory at runtime, but we believe it's more robust to have the program manage its own state as much as possible. Thanks, Dan
At Mon, 8 Jun 2015 05:52:18 -0700, Dan Nicholson wrote: > > On Jun 8, 2015 4:38 AM, "Takashi Iwai" <tiwai@suse.de> wrote: > > > > At Fri, 5 Jun 2015 15:00:47 -0700, > > Dan Nicholson wrote: > > > > > > Try to create the directory for the state file when saving so we don't > > > depend on it being created ahead of time. This only checks for failures > > > on existing directories and doesn't try to create the leading > > > directories or workaround any other errors. This should catch the common > > > case where /var/lib exists, but /var/lib/alsa doesn't. > > > > I don't think it's the role of alsactl. It saves a file on the > > certain directory. If it doesn't exist, it's a failure of the > > installed package. > > Sure, that's understandable, but there's a couple reasons I think this is > helpful addition. > > First, if no path is supplied, store will save to /var/lib/alsa. So, it's > not as of the user has supplied a path it didn't setup correctly. It would > be nice if alsactl worked out of the box without additional integration by > packagers. For that, a safer way would be to create /var/lib/alsa in the installation. > Second, my real motivation for fixing this is to support stateless type of > systems that come with a clean /var. At Endless we're using ostree. The OS > is composed by Debian packages, and indeed alsa-utils is setup to create > /var/lib/alsa on install. However, to use the same OS snapshot for all > users, the contents of /var are stripped since they represent local system > state. We can certainly add a method for creating the directory at runtime, > but we believe it's more robust to have the program manage its own state as > much as possible. Why not specifying the proper directory via alsactl -f option for user? Takashi
At Mon, 08 Jun 2015 15:03:21 +0200, Takashi Iwai wrote: > > At Mon, 8 Jun 2015 05:52:18 -0700, > Dan Nicholson wrote: > > > > On Jun 8, 2015 4:38 AM, "Takashi Iwai" <tiwai@suse.de> wrote: > > > > > > At Fri, 5 Jun 2015 15:00:47 -0700, > > > Dan Nicholson wrote: > > > > > > > > Try to create the directory for the state file when saving so we don't > > > > depend on it being created ahead of time. This only checks for failures > > > > on existing directories and doesn't try to create the leading > > > > directories or workaround any other errors. This should catch the common > > > > case where /var/lib exists, but /var/lib/alsa doesn't. > > > > > > I don't think it's the role of alsactl. It saves a file on the > > > certain directory. If it doesn't exist, it's a failure of the > > > installed package. > > > > Sure, that's understandable, but there's a couple reasons I think this is > > helpful addition. > > > > First, if no path is supplied, store will save to /var/lib/alsa. So, it's > > not as of the user has supplied a path it didn't setup correctly. It would > > be nice if alsactl worked out of the box without additional integration by > > packagers. > > For that, a safer way would be to create /var/lib/alsa in the > installation. > > > Second, my real motivation for fixing this is to support stateless type of > > systems that come with a clean /var. At Endless we're using ostree. The OS > > is composed by Debian packages, and indeed alsa-utils is setup to create > > /var/lib/alsa on install. However, to use the same OS snapshot for all > > users, the contents of /var are stripped since they represent local system > > state. We can certainly add a method for creating the directory at runtime, > > but we believe it's more robust to have the program manage its own state as > > much as possible. > > Why not specifying the proper directory via alsactl -f option for > user? Also, your patch seems to care only the top directory. For example, if alsactl is configured to use /var/lib/alsa/more/deep/dir, it won't work. thanks, Takashi
diff --git a/alsactl/state.c b/alsactl/state.c index 3908ec4..8ca3d6e 100644 --- a/alsactl/state.c +++ b/alsactl/state.c @@ -27,6 +27,9 @@ #include <stdio.h> #include <assert.h> #include <errno.h> +#include <string.h> +#include <sys/stat.h> +#include <sys/types.h> #include <alsa/asoundlib.h> #include "alsactl.h" @@ -1544,6 +1547,7 @@ int save_state(const char *file, const char *cardname) snd_output_t *out; int stdio; char *nfile = NULL; + char *filedir = NULL; int lock_fd = -EINVAL; err = snd_config_top(&config); @@ -1553,6 +1557,9 @@ int save_state(const char *file, const char *cardname) } stdio = !strcmp(file, "-"); if (!stdio) { + char *tmp; + mode_t mode = S_IRWXU | S_IRWXG | S_IRWXO; + nfile = malloc(strlen(file) + 5); if (nfile == NULL) { error("No enough memory..."); @@ -1561,6 +1568,21 @@ int save_state(const char *file, const char *cardname) } strcpy(nfile, file); strcat(nfile, ".new"); + filedir = strdup(file); + if (filedir == NULL) { + error("Not enough memory..."); + err = -ENOMEM; + goto out; + } + tmp = strrchr(filedir, '/'); + if (tmp && tmp != filedir) { + *tmp = '\0'; + if (mkdir(filedir, mode) != 0 && errno != EEXIST) { + error("Could not create directory %s: %s", + filedir, strerror(errno)); + goto out; + } + } lock_fd = state_lock(file, 10); if (lock_fd < 0) { err = lock_fd; @@ -1640,6 +1662,7 @@ out: if (!stdio && lock_fd >= 0) state_unlock(lock_fd, file); free(nfile); + free(filedir); snd_config_delete(config); snd_config_update_free_global(); return err;
Try to create the directory for the state file when saving so we don't depend on it being created ahead of time. This only checks for failures on existing directories and doesn't try to create the leading directories or workaround any other errors. This should catch the common case where /var/lib exists, but /var/lib/alsa doesn't. Signed-off-by: Dan Nicholson <nicholson@endlessm.com> --- alsactl/state.c | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+)