From: "Benjamin Marzinski" <bmarzins@redhat.com>
To: Martin Wilck <mwilck@suse.com>
Cc: dm-devel@redhat.com
Subject: Re: [PATCH 03/21] libmultipath/checkers: replace message by msgid
Date: Thu, 25 Oct 2018 15:57:27 -0500 [thread overview]
Message-ID: <20181025205727.GL7100@octiron.msp.redhat.com> (raw)
In-Reply-To: <20181011222707.3631-4-mwilck@suse.com>
On Fri, Oct 12, 2018 at 12:26:49AM +0200, Martin Wilck wrote:
> Replace the character array "message" in struct checker with
> a "message ID" field.
>
> The generic checker code defines a couple of standard message IDs
> and corresponding messages. Checker-specific message IDs start
> at CHECKER_FIRST_MSG. Checkers that implement specific message
> IDs must provide a table for converting the IDs into actual log
> messages.
>
> This simplifies the checker data structure and the handling of
> checker messages in the critical checker code path. It comes at
> the cost that only constant message strings are supported. It
> turns out that only a single checker log message (in the emc_clariion
> checker) was dynamically generated, and the missing information can
> be provided with a standard condlog message.
>
> Follow-up patches implement this for the existing checkers.
I have some cosmetic nits that I'll add inline, but none of them are
real problems so,
Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>
>
> Signed-off-by: Martin Wilck <mwilck@suse.com>
> ---
> libmultipath/checkers.c | 70 +++++++++++++++++++++++++++++++++++------
> libmultipath/checkers.h | 39 +++++++++++++++++++----
> 2 files changed, 94 insertions(+), 15 deletions(-)
>
> diff --git a/libmultipath/checkers.c b/libmultipath/checkers.c
> index 0bacc864..359791ff 100644
> --- a/libmultipath/checkers.c
> +++ b/libmultipath/checkers.c
> @@ -141,6 +141,22 @@ struct checker * add_checker (char *multipath_dir, char * name)
> if (!c->free)
> goto out;
>
> + c->msgtable_size = 0;
> + c->msgtable = dlsym(c->handle, "libcheck_msgtable");
> +
> + if (c->msgtable != NULL) {
> + const char **p;
> +
> + for (p = c->msgtable;
> + *p && (p - c->msgtable) < CHECKER_MSGTABLE_SIZE; p++)
> + /* nothing */;
> +
> + c->msgtable_size = p - c->msgtable;
> + } else
> + c->msgtable_size = 0;
> + condlog(3, "checker %s: message table size = %d",
> + c->name, c->msgtable_size);
> +
> done:
> c->fd = -1;
> c->sync = 1;
> @@ -222,16 +238,16 @@ int checker_check (struct checker * c, int path_state)
> if (!c)
> return PATH_WILD;
>
> - c->message[0] = '\0';
> + c->msgid = CHECKER_MSGID_NONE;
> if (c->disable) {
> - MSG(c, "checker disabled");
> + c->msgid = CHECKER_MSGID_DISABLED;
> return PATH_UNCHECKED;
> }
> if (!strncmp(c->name, NONE, 4))
> return path_state;
>
> if (c->fd < 0) {
> - MSG(c, "no usable fd");
> + c->msgid = CHECKER_MSGID_NO_FD;
> return PATH_WILD;
> }
> r = c->check(c);
> @@ -248,25 +264,59 @@ int checker_selected (struct checker * c)
> return (c->check) ? 1 : 0;
> }
>
> -char * checker_name (struct checker * c)
> +const char *checker_name(const struct checker *c)
> {
> if (!c)
> return NULL;
> return c->name;
> }
>
> -char * checker_message (struct checker * c)
> +static const char *generic_msg[CHECKER_LAST_GENERIC_MSGID] = {
> + [CHECKER_MSGID_NONE] = "",
> + [CHECKER_MSGID_DISABLED] = " is disabled",
> + [CHECKER_MSGID_NO_FD] = " has no usable fd",
> + [CHECKER_MSGID_INVALID] = " provided invalid message id",
> + [CHECKER_MSGID_UP] = " reports path is up",
> + [CHECKER_MSGID_DOWN] = " reports path is down",
> + [CHECKER_MSGID_GHOST] = " reports path is ghost",
> +};
> +
> +static const char *_checker_message(const struct checker *c)
> {
> - if (!c)
> + int id;
> +
> + if (!c || c->msgid < 0 ||
> + (c->msgid >= CHECKER_LAST_GENERIC_MSGID &&
> + c->msgid < CHECKER_FIRST_MSGID))
> return NULL;
> - return c->message;
> +
> + if (c->msgid < CHECKER_LAST_GENERIC_MSGID)
> + return generic_msg[c->msgid];
> +
> + id = c->msgid - CHECKER_FIRST_MSGID;
> + if (id < c->msgtable_size)
> + return c->msgtable[id];
> + return NULL;
> +}
> +
> +const char *checker_message(const struct checker *c)
> +{
> + static char buf[CHECKER_MSG_LEN];
> + const char *msg = _checker_message(c);
> +
> + if (msg == NULL || *msg == '\0')
> + *buf = '\0';
> + else
> + snprintf(buf, sizeof(buf), "%s checker%s",
> + c->name, msg);
> + return buf;
> }
Since we are returning a pointer to a static buffer, it is theoretically
possible for multiple threads to be updating the buffer at the same
time. In reality, we are always calling checker_message with the vecs
lock held, seeing as we need to guarantee that the path we are checking
doesn't get deleted from pathvec while we are using it. However, it is
easy to imagine future code calling checker_message for a path that
hasn't been added to pathvec, and thus couldn't be deleted by another
thread. In this case we wouldn't need the vecs lock, except to guarantee
that we are the only thread using the static buffer. I don't really have
a problem with keeping the static buffer, but we should add a comment
that checker_message needs to be called with the vecs lock held.
> void checker_clear_message (struct checker *c)
> {
> if (!c)
> return;
> - c->message[0] = '\0';
> + c->msgid = CHECKER_MSGID_NONE;
> }
>
> void checker_get (char *multipath_dir, struct checker * dst, char * name)
> @@ -288,10 +338,12 @@ void checker_get (char *multipath_dir, struct checker * dst, char * name)
> dst->fd = src->fd;
> dst->sync = src->sync;
> strncpy(dst->name, src->name, CHECKER_NAME_LEN);
> - strncpy(dst->message, src->message, CHECKER_MSG_LEN);
> + dst->msgid = src->msgid;
It looks odd setting dst->msgid to src->msgid, when we've never set
src->msgid. I realize that CHECKER_MSGID_NONE = 0, and we zero out the
checker when we alloc is, so src->msgid is set to CHECKER_MSGID_NONE.
But since we are always referring to msgid as a named constant, it would
be clearer if we explicitly set it to CHECKER_MSGID_NONE in add_checker
(even though it is unnecessary). This is purely cosmetic.
> dst->check = src->check;
> dst->init = src->init;
> dst->free = src->free;
> + dst->msgtable = src->msgtable;
> + dst->msgtable_size = src->msgtable_size;
> dst->handle = NULL;
> src->refcount++;
> }
> diff --git a/libmultipath/checkers.h b/libmultipath/checkers.h
> index 7b18a1ac..b48f4dc2 100644
> --- a/libmultipath/checkers.h
> +++ b/libmultipath/checkers.h
> @@ -97,6 +97,22 @@ enum path_check_state {
> #define CHECKER_DEV_LEN 256
> #define LIB_CHECKER_NAMELEN 256
>
> +/*
> + * Generic message IDs for use in checkers.
> + */
> +enum {
> + CHECKER_MSGID_NONE = 0,
> + CHECKER_MSGID_DISABLED,
> + CHECKER_MSGID_NO_FD,
> + CHECKER_MSGID_INVALID,
> + CHECKER_MSGID_UP,
> + CHECKER_MSGID_DOWN,
> + CHECKER_MSGID_GHOST,
> + CHECKER_LAST_GENERIC_MSGID,
I prefer CHECKER_GENERIC_MSGTABLE_SIZE to CHECKER_LAST_GENERIC_MSGID.
"char *generic_msg[CHECKER_LAST_GENERIC_MSGID]" just looks like it's an
off-by-one error when you see it. Again, this is clearly cosmetic. It
just made me worried when I first saw it.
> + CHECKER_FIRST_MSGID = 100, /* lowest msgid for checkers */
> + CHECKER_MSGTABLE_SIZE = 100, /* max msg table size for checkers */
> +};
> +
> struct checker {
> struct list_head node;
> void *handle;
> @@ -106,7 +122,7 @@ struct checker {
> unsigned int timeout;
> int disable;
> char name[CHECKER_NAME_LEN];
> - char message[CHECKER_MSG_LEN]; /* comm with callers */
> + short msgid; /* checker-internal extra status */
> void * context; /* store for persistent data */
> void ** mpcontext; /* store for persistent data shared
> multipath-wide. Use MALLOC if
> @@ -114,10 +130,10 @@ struct checker {
> int (*check)(struct checker *);
> int (*init)(struct checker *); /* to allocate the context */
> void (*free)(struct checker *); /* to free the context */
> + const char**msgtable;
> + short msgtable_size;
> };
>
> -#define MSG(c, fmt, args...) snprintf((c)->message, CHECKER_MSG_LEN, fmt, ##args);
> -
> char * checker_state_name (int);
> int init_checkers (char *);
> void cleanup_checkers (void);
> @@ -134,14 +150,25 @@ void checker_enable (struct checker *);
> void checker_disable (struct checker *);
> int checker_check (struct checker *, int);
> int checker_selected (struct checker *);
> -char * checker_name (struct checker *);
> -char * checker_message (struct checker *);
> +const char *checker_name (const struct checker *);
> +const char *checker_message (const struct checker *);
> void checker_clear_message (struct checker *c);
> void checker_get (char *, struct checker *, char *);
>
> -/* Functions exported by path checker dynamic libraries (.so) */
> +/* Prototypes for symbols exported by path checker dynamic libraries (.so) */
> int libcheck_check(struct checker *);
> int libcheck_init(struct checker *);
> void libcheck_free(struct checker *);
> +/*
> + * msgid => message map.
> + *
> + * It only needs to be provided if the checker defines specific
> + * message IDs.
> + * Message IDs available to checkers start at CHECKER_FIRST_MSG.
> + * The msgtable array is 0-based, i.e. msgtable[0] is the message
> + * for msgid == __CHECKER_FIRST_MSG.
> + * The table ends with a NULL element.
> + */
> +extern const char *libcheck_msgtable[];
>
> #endif /* _CHECKERS_H */
> --
> 2.19.0
next prev parent reply other threads:[~2018-10-25 20:57 UTC|newest]
Thread overview: 47+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-10-11 22:26 [PATCH 00/21] libmultipath: checkers overhaul Martin Wilck
2018-10-11 22:26 ` [PATCH 01/21] libmultipath: fix use of uninitialized memory in write() Martin Wilck
2018-10-25 20:31 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 02/21] libmultipath: fix memory leaks from scandir() use Martin Wilck
2018-10-25 20:31 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 03/21] libmultipath/checkers: replace message by msgid Martin Wilck
2018-10-25 20:57 ` Benjamin Marzinski [this message]
2018-10-25 21:36 ` Martin Wilck
2018-10-11 22:26 ` [PATCH 04/21] libmultipath/checkers: cciss_tur: use message id Martin Wilck
2018-10-25 20:57 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 05/21] libmultipath/checkers: directio: " Martin Wilck
2018-10-25 20:58 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 06/21] libmultipath/checkers: emc_clariion: " Martin Wilck
2018-10-25 20:58 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 07/21] libmultipath/checkers: hp_sw: " Martin Wilck
2018-10-25 20:58 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 08/21] libmultipath/checkers: rdac: " Martin Wilck
2018-10-25 20:59 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 09/21] libmultipath/checkers: readsector0: " Martin Wilck
2018-10-25 21:02 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 10/21] libmultipath/checkers: tur: " Martin Wilck
2018-10-25 21:32 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 11/21] multipathd: improve checker message logging Martin Wilck
2018-10-25 21:32 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 12/21] libmultipath/checkers: support unsupported paths Martin Wilck
2018-10-25 21:33 ` Benjamin Marzinski
2018-10-11 22:26 ` [PATCH 13/21] libmultipath: clariion checker: leave unsupported paths alone Martin Wilck
2018-10-25 21:41 ` Benjamin Marzinski
2018-10-26 9:01 ` Martin Wilck
2018-10-11 22:27 ` [PATCH 14/21] libmultipath: hp_sw " Martin Wilck
2018-10-25 21:42 ` Benjamin Marzinski
2018-10-11 22:27 ` [PATCH 15/21] libmultipath: rdac " Martin Wilck
2018-10-25 21:42 ` Benjamin Marzinski
2018-10-11 22:27 ` [PATCH 16/21] libmultipath: tur " Martin Wilck
2018-10-25 21:42 ` Benjamin Marzinski
2018-10-11 22:27 ` [PATCH 17/21] libmultipath: pathinfo: don't blank wwid if checker fails Martin Wilck
2018-10-25 21:50 ` Benjamin Marzinski
2018-10-26 9:16 ` Martin Wilck
2018-10-26 15:11 ` Benjamin Marzinski
2018-10-11 22:27 ` [PATCH 18/21] multipathd: check_path: improve logging for "unusable path" case Martin Wilck
2018-10-25 21:50 ` Benjamin Marzinski
2018-10-11 22:27 ` [PATCH 19/21] libmultipath: coalesce_paths: improve logging of orphaned paths Martin Wilck
2018-10-25 21:50 ` Benjamin Marzinski
2018-10-11 22:27 ` [PATCH 20/21] libmultipath: sync_map_state: log failing paths Martin Wilck
2018-10-25 21:52 ` Benjamin Marzinski
2018-10-11 22:27 ` [PATCH 21/21] libmultipath/checkers: cleanup class/instance model Martin Wilck
2018-10-25 21:52 ` Benjamin Marzinski
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20181025205727.GL7100@octiron.msp.redhat.com \
--to=bmarzins@redhat.com \
--cc=dm-devel@redhat.com \
--cc=mwilck@suse.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox