Skip to contentSkip to conversation

Fixes from clang

PATCH New Zdenek Kabelac 5 participants 70 messages Oct 25, 2010 – Oct 26, 2010

lore
Share & tools
lore
#0 lore raw

cover Fixes from clang

Patchset combines code cleanups for compiler warnings and code fixes found by static analyzer scan-build.

Potentional NULL pointer dereferencies are mostly in very unusual code execution paths. I've tried to order them on their importance - the later ones are not event trigerable as surrounding code should not be to pass problematic values - but checks seems to be simple enough to keep static analysis happy.

Last patches in this set add some instrumentation to function calls that might allow better code optimization or better static analysis.

Zdenek Kabelac (30):
Fix clang warning for ntohl(*((uint32_t *)buf)) Use dm_pool_grow_object embeded strlen for 0 delta Check (type) is not NULL before access
Fix NULL pointer dereference
Print vg_name and do not to access vg->name Ensure we always have origin defined Ensure seg is nonnull
Ensure first is not NULL
Reuse result of previous strchr.
Fix potential NULL pointer dereference
Fix theoretical usage of NULL pointer dereference Check for NULL pointer
Eliminate uninitialized_var
Fix missing initilisation to 0
Function pull_state cannot work with NULL buffer Fix void* pointer arithmetic
Use const pvid for device_from_pvid()
Use const pointer for return value of dm_basename Fix constness warning
Fix constness warning
Fix constness warning
Fix constness warning
Using const config node
bufused is assigned 0 in preceding source line Proper C declaration
Use static
Instrument compiler about code unreachability Skip uninitialized macro for scan-build Add some __attribute__ instrumentation __attribute__((nonnull(1)))

 daemons/clvmd/clvmd.c                 |    7 +++++--
 daemons/cmirrord/cluster.c            |    2 +-
 daemons/cmirrord/functions.c          |    4 +++-
 daemons/dmeventd/dmeventd.c           |   21 +++++++++++----------
 daemons/dmeventd/libdevmapper-event.c |   20 ++++++++++----------
 lib/activate/dev_manager.c            |    2 +-
 lib/cache/lvmcache.c                  |   10 +++++-----
 lib/cache/lvmcache.h                  |    2 +-
 lib/config/config.c                   |    2 +-
 lib/device/dev-cache.h                |    2 +-
 lib/device/dev-md.c                   |    3 ++-
 lib/format1/format1.c                 |    2 +-
 lib/format_text/archive.c             |    4 ++--
 lib/format_text/format-text.c         |    2 +-
 lib/log/log.c                         |    2 +-
 lib/metadata/lv_manip.c               |    2 +-
 lib/metadata/metadata.c               |    6 +++---
 lib/misc/crc.c                        |    6 +++---
 lib/misc/util.h                       |    4 ++++
 lib/mm/memlock.c                      |    4 ++--
 lib/report/report.c                   |    6 +++---
 lib/striped/striped.c                 |    2 +-
 libdm/libdevmapper.h                  |   28 ++++++++++++++--------------
 libdm/libdm-report.c                  |    7 ++++---
 libdm/libdm-string.c                  |    6 +++---
 libdm/mm/pool-fast.c                  |    8 ++++----
 libdm/regex/ttree.c                   |    7 +++++--
 tools/dmsetup.c                       |    2 +-
 tools/lvcreate.c                      |    4 ++++
 tools/lvm.c                           |    6 +++---
 tools/lvmcmdline.c                    |    2 +-
 tools/pvresize.c                      |    2 +-
 tools/reporter.c                      |    2 +-
 tools/toollib.c                       |    2 +-
 34 files changed, 105 insertions(+), 86 deletions(-)

--
1.7.3.1

#1 lore raw

PATCH 1/30 Fix clang warning for ntohl(*((uint32_t *)buf))

We cast (char*) to (uint32_t*) that changes alignment requierements. For our case the code has been correct as alloca() returns properly aligned buffer, however this patch make it cleaner and more readable and avoids warning generation.

FIXME: code duplication

---
 daemons/dmeventd/dmeventd.c           |   17 +++++++++--------
 daemons/dmeventd/libdevmapper-event.c |   20 ++++++++++----------
 2 files changed, 19 insertions(+), 18 deletions(-)
daemons/dmeventd/dmeventd.c +9 -8
index 639f200..d179879 100644
@@ -1300,9 +1300,9 @@ static int _client_read(struct dm_event_fifos *fifos,
 	unsigned bytes = 0;
 	int ret = 0;
 	fd_set fds;
-	int header = 1;
 	size_t size = 2 * sizeof(uint32_t);	/* status + size */
-	char *buf = alloca(size);
+	uint32_t *header = alloca(size);
+	char *buf = (char *)header;
 
 	msg->data = NULL;
 
@@ -1326,9 +1326,9 @@ static int _client_read(struct dm_event_fifos *fifos,
 
 		ret = read(fifos->client, buf + bytes, size - bytes);
 		bytes += ret > 0 ? ret : 0;
-		if (bytes == 2 * sizeof(uint32_t) && header) {
-			msg->cmd = ntohl(*((uint32_t *) buf));
-			msg->size = ntohl(*((uint32_t *) buf + 1));
+		if (header && (bytes == 2 * sizeof(uint32_t))) {
+			msg->cmd = ntohl(header[0]);
+			msg->size = ntohl(header[1]);
 			buf = msg->data = dm_malloc(msg->size);
 			size = msg->size;
 			bytes = 0;
@@ -1356,10 +1356,11 @@ static int _client_write(struct dm_event_fifos *fifos,
 	fd_set fds;
 
 	size_t size = 2 * sizeof(uint32_t) + msg->size;
-	char *buf = alloca(size);
+	uint32_t *header = alloca(size);
+	char *buf = (char *)header;
 
-	*((uint32_t *)buf) = htonl(msg->cmd);
-	*((uint32_t *)buf + 1) = htonl(msg->size);
+	header[0] = htonl(msg->cmd);
+	header[1] = htonl(msg->size);
 	if (msg->data)
 		memcpy(buf + 2 * sizeof(uint32_t), msg->data, msg->size);
 
daemons/dmeventd/libdevmapper-event.c +10 -10
index abbb968..bc8ad99 100644
@@ -230,8 +230,8 @@ static int _daemon_read(struct dm_event_fifos *fifos,
 	fd_set fds;
 	struct timeval tval = { 0, 0 };
 	size_t size = 2 * sizeof(uint32_t);	/* status + size */
-	char *buf = alloca(size);
-	int header = 1;
+	uint32_t *header = alloca(size);
+	char *buf = (char *)header;
 
 	while (bytes < size) {
 		for (i = 0, ret = 0; (i < 20) && (ret < 1); i++) {
@@ -262,9 +262,9 @@ static int _daemon_read(struct dm_event_fifos *fifos,
 		}
 
 		bytes += ret;
-		if (bytes == 2 * sizeof(uint32_t) && header) {
-			msg->cmd = ntohl(*((uint32_t *)buf));
-			msg->size = ntohl(*((uint32_t *)buf + 1));
+		if (header && (bytes == 2 * sizeof(uint32_t))) {
+			msg->cmd = ntohl(header[0]);
+			msg->size = ntohl(header[1]);
 			buf = msg->data = dm_malloc(msg->size);
 			size = msg->size;
 			bytes = 0;
@@ -288,12 +288,13 @@ static int _daemon_write(struct dm_event_fifos *fifos,
 	fd_set fds;
 
 	size_t size = 2 * sizeof(uint32_t) + msg->size;
-	char *buf = alloca(size);
+	uint32_t *header = alloca(size);
+	char *buf = (char *)header;
 	char drainbuf[128];
 	struct timeval tval = { 0, 0 };
 
-	*((uint32_t *)buf) = htonl(msg->cmd);
-	*((uint32_t *)buf + 1) = htonl(msg->size);
+	header[0] = htonl(msg->cmd);
+	header[1] = htonl(msg->size);
 	memcpy(buf + 2 * sizeof(uint32_t), msg->data, msg->size);
 
 	/* drain the answer fifo */
@@ -323,8 +324,7 @@ static int _daemon_write(struct dm_event_fifos *fifos,
 			}
 		} while (ret < 1);
 
-		ret = write(fifos->client, ((char *) buf) + bytes,
-			    size - bytes);
+		ret = write(fifos->client, buf + bytes, size - bytes);
 		if (ret < 0) {
 			if ((errno == EINTR) || (errno == EAGAIN))
 				continue;
Signature · 1 line
1.7.3.1
#3 lore raw

PATCH 2/30 Use dm_pool_grow_object embeded strlen for 0 delta

In reply to Zdenek Kabelac

Use internal strlen for 0

---
 lib/config/config.c       |    2 +-
 lib/format_text/archive.c |    4 ++--
 lib/metadata/lv_manip.c   |    2 +-
 lib/metadata/metadata.c   |    4 ++--
 lib/report/report.c       |    6 +++---
 tools/lvmcmdline.c        |    2 +-
 6 files changed, 10 insertions(+), 10 deletions(-)
lib/config/config.c +1 -1
index 2b38663..e64ec86 100644
@@ -387,7 +387,7 @@ static int _line_append(struct output_line *outline, const char *fmt, ...)
 	}
 	va_end(ap);
 
-	if (!dm_pool_grow_object(outline->mem, &buf[0], strlen(buf))) {
+	if (!dm_pool_grow_object(outline->mem, buf, 0)) {
 		log_error("dm_pool_grow_object failed for config line");
 		return 0;
 	}
lib/format_text/archive.c +2 -2
index bf16da5..246350b 100644
@@ -110,9 +110,9 @@ static void _insert_archive_file(struct dm_list *head, struct archive_file *b)
 static char *_join_file_to_dir(struct dm_pool *mem, const char *dir, const char *name)
 {
 	if (!dm_pool_begin_object(mem, 32) ||
-	    !dm_pool_grow_object(mem, dir, strlen(dir)) ||
+	    !dm_pool_grow_object(mem, dir, 0) ||
 	    !dm_pool_grow_object(mem, "/", 1) ||
-	    !dm_pool_grow_object(mem, name, strlen(name)) ||
+	    !dm_pool_grow_object(mem, name, 0) ||
 	    !dm_pool_grow_object(mem, "\0", 1))
 		return_NULL;
 
lib/metadata/lv_manip.c +1 -1
index e4349cc..0ea7dcf 100644
@@ -665,7 +665,7 @@ static int _log_parallel_areas(struct dm_pool *mem, struct dm_list *parallel_are
 		}
 
 		dm_list_iterate_items(pvl, &spvs->pvs) {
-			if (!dm_pool_grow_object(mem, pv_dev_name(pvl->pv), strlen(pv_dev_name(pvl->pv)))) {
+			if (!dm_pool_grow_object(mem, pv_dev_name(pvl->pv), 0)) {
 				log_error("dm_pool_grow_object failed");
 				dm_pool_abandon_object(mem);
 				return 0;
lib/metadata/metadata.c +2 -2
index 4d11ac4..2da4085 100644
@@ -1078,7 +1078,7 @@ static dm_bitset_t _bitset_with_random_bits(struct dm_pool *mem, uint32_t num_bi
 			dm_pool_free(mem, bs);
                 	return NULL;
 		}
-		if (!dm_pool_grow_object(mem, buf, strlen(buf))) {
+		if (!dm_pool_grow_object(mem, buf, 0)) {
 			log_error("Failed to generate list of random bits.");
 			dm_pool_free(mem, bs);
                 	return NULL;
@@ -3970,7 +3970,7 @@ char *tags_format_and_copy(struct dm_pool *mem, const struct dm_list *tags)
 	}
 
 	dm_list_iterate_items(sl, tags) {
-		if (!dm_pool_grow_object(mem, sl->str, strlen(sl->str)) ||
+		if (!dm_pool_grow_object(mem, sl->str, 0) ||
 		    (sl->list.n != tags && !dm_pool_grow_object(mem, ",", 1))) {
 			log_error("dm_pool_grow_object failed");
 			return NULL;
lib/report/report.c +3 -3
index 7a90b4d..b26a1e3 100644
@@ -85,7 +85,7 @@ static int _format_pvsegs(struct dm_pool *mem, struct dm_report_field *field,
 			extent = 0;
 		}
 
-		if (!dm_pool_grow_object(mem, name, strlen(name))) {
+		if (!dm_pool_grow_object(mem, name, 0)) {
 			log_error("dm_pool_grow_object failed");
 			return 0;
 		}
@@ -97,7 +97,7 @@ static int _format_pvsegs(struct dm_pool *mem, struct dm_report_field *field,
 			log_error("Extent number dm_snprintf failed");
 			return 0;
 		}
-		if (!dm_pool_grow_object(mem, extent_str, strlen(extent_str))) {
+		if (!dm_pool_grow_object(mem, extent_str, 0)) {
 			log_error("dm_pool_grow_object failed");
 			return 0;
 		}
@@ -108,7 +108,7 @@ static int _format_pvsegs(struct dm_pool *mem, struct dm_report_field *field,
 				log_error("Extent number dm_snprintf failed");
 				return 0;
 			}
-			if (!dm_pool_grow_object(mem, extent_str, strlen(extent_str))) {
+			if (!dm_pool_grow_object(mem, extent_str, 0)) {
 				log_error("dm_pool_grow_object failed");
 				return 0;
 			}
tools/lvmcmdline.c +1 -1
index b29acfb..06873a0 100644
@@ -986,7 +986,7 @@ static const char *_copy_command_line(struct cmd_context *cmd, int argc, char **
 		if (space && !dm_pool_grow_object(cmd->mem, "'", 1))
 			goto_bad;
 
-		if (!dm_pool_grow_object(cmd->mem, argv[i], strlen(argv[i])))
+		if (!dm_pool_grow_object(cmd->mem, argv[i], 0))
 			goto_bad;
 
 		if (space && !dm_pool_grow_object(cmd->mem, "'", 1))
Signature · 1 line
1.7.3.1
#7 lore raw

PATCH 3/30 Check (type) is not NULL before access

In reply to Zdenek Kabelac

clang Logic error Dereference of null pointer

---
 libdm/libdm-report.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)
libdm/libdm-report.c +4 -3
index 7631e21..28b27b2 100644
@@ -273,9 +273,10 @@ static void _display_fields(struct dm_report *rh)
 			log_warn("%*.*s", (int) strlen(desc) + 7,
 				 (int) strlen(desc) + 7,
 				 "-------------------------------------------------------------------------------");
-			log_warn("  %sall%-*s - %s", type->prefix,
-				 (int) (id_len - 3 - strlen(type->prefix)), "",
-				 "All fields in this section.");
+			if (type)
+				log_warn("  %sall%-*s - %s", type->prefix,
+					 (int) (id_len - 3 - strlen(type->prefix)), "",
+					 "All fields in this section.");
 		}
 
 		/* FIXME Add line-wrapping at terminal width (or 80 cols) */
Signature · 1 line
1.7.3.1
#9 lore raw

PATCH 4/30 Fix NULL pointer dereference

In reply to Zdenek Kabelac

vg pointer is not yet initialized in this place and vgname must be used.

---
 lib/format_text/format-text.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
lib/format_text/format-text.c +1 -1
index eee0114..0102a15 100644
@@ -507,7 +507,7 @@ static struct volume_group *_vg_read_raw_area(struct format_instance *fid,
 
 	if (wrap > rlocn->offset) {
 		log_error("VG %s metadata too large for circular buffer",
-			  vg->name);
+			  vgname);
 		goto out;
 	}
 
Signature · 1 line
1.7.3.1
#11 lore raw

PATCH 5/30 Print vg_name and do not to access vg->name

In reply to Zdenek Kabelac

clang Logic error Dereference of null pointer
Replacing pointer access with vg_name.

---
 tools/pvresize.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/pvresize.c +1 -1
index ccd025a..af499e5 100644
@@ -148,7 +148,7 @@ static int _pv_resize_single(struct cmd_context *cmd,
 	if (!is_orphan_vg(vg_name)) {
 		if (!vg_write(vg) || !vg_commit(vg)) {
 			log_error("Failed to store physical volume \"%s\" in "
-				  "volume group \"%s\"", pv_name, vg->name);
+				  "volume group \"%s\"", pv_name, vg_name);
 			goto out;
 		}
 		backup(vg);
Signature · 1 line
1.7.3.1
#13 lore raw

PATCH 6/30 Ensure we always have origin defined

In reply to Zdenek Kabelac

clang Logic error Dereference of null pointer
Ensure the code path could not use origin == NULL because of internal error in code.

---
 tools/lvcreate.c |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)
tools/lvcreate.c +4 -0
index de5b59c..a9e3c2d 100644
@@ -179,6 +179,10 @@ static int _update_extents_params(struct volume_group *vg,
 					  lp->origin);
 				return 0;
 			}
+			if (!origin) {
+				log_error(INTERNAL_ERROR "Couldn't find origin volume.");
+				return 0;
+			}
 			lp->extents = lp->extents * origin->le_count / 100;
 			break;
 		case PERCENT_NONE:
Signature · 1 line
1.7.3.1
#15 lore raw

PATCH 7/30 Ensure seg is nonnull

In reply to Zdenek Kabelac

clang Logic error Dereference of null pointer
Make sure we do not try to use NULL seg pointer.

---
 lib/activate/dev_manager.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
lib/activate/dev_manager.c +1 -1
index aab0c9a..216b042 100644
@@ -615,7 +615,7 @@ int dev_manager_transient(struct dev_manager *dm, struct logical_volume *lv)
 		if (!type || !params)
 			continue;
 
-		if (seg->segtype->ops->check_transient_status &&
+		if (seg && seg->segtype->ops->check_transient_status &&
 		    !seg->segtype->ops->check_transient_status(seg, params))
 			goto_out;
 
Signature · 1 line
1.7.3.1
#17 lore raw

PATCH 8/30 Ensure first is not NULL

In reply to Zdenek Kabelac

clang Logic error Dereference of null pointer

For empty lists make sure the NULL first pointer is not dereferenced.

---
 lib/format1/format1.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
lib/format1/format1.c +1 -1
index 0dcec05..0efde31 100644
@@ -103,7 +103,7 @@ static int _check_vgs(struct dm_list *pvs, struct volume_group *vg)
 	}
 
 	/* On entry to fn, list known to be non-empty */
-	if (pv_count != first->vgd.pv_cur) {
+	if (first && (pv_count != first->vgd.pv_cur)) {
 		log_error("%d PV(s) found for VG %s: expected %d",
 			  pv_count, first->pvd.vg_name, first->vgd.pv_cur);
 		vg->status |= PARTIAL_VG;
Signature · 1 line
1.7.3.1
#19 lore raw

PATCH 9/30 Reuse result of previous strchr.

In reply to Zdenek Kabelac

clang Argument with 'nonnull' attribute passed null Reuse result of the last strchr call - it also makes sure, 'st' point is nonnull for next strchr call.

---
 tools/toollib.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/toollib.c +1 -1
index 02c4962..5adc160 100644
@@ -869,7 +869,7 @@ const char *extract_vgname(struct cmd_context *cmd, const char *lv_name)
 			while (*st == '/')
 				st++;
 
-		if (!strchr(vg_name, '/') || strchr(st, '/')) {
+		if (!st || strchr(st, '/')) {
 			log_error("\"%s\": Invalid path for Logical Volume",
 				  lv_name);
 			return 0;
Signature · 1 line
1.7.3.1
#21 lore raw

PATCH 10/30 Fix potential NULL pointer dereference

In reply to Zdenek Kabelac

Makes clang happier as it covers all code paths and avoids NULL pointer dereference through com pointer (which is NULL by default static initialisation).

---
 tools/lvm.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)
tools/lvm.c +3 -3
index 60f9a79..9cac23e 100644
@@ -85,11 +85,11 @@ static char *_list_args(const char *text, int state)
 				break;
 			}
 		}
-
-		if (!com)
-			return NULL;
 	}
 
+	if (!com)
+		return NULL;
+
 	/* Short form arguments */
 	if (len < 3) {
 		while (match_no < com->num_args) {
Signature · 1 line
1.7.3.1
#23 lore raw

PATCH 11/30 Fix theoretical usage of NULL pointer dereference

In reply to Zdenek Kabelac

clang seems to be happier with this check.

---
 libdm/regex/ttree.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)
libdm/regex/ttree.c +3 -1
index ec97c98..0ad40bd 100644
@@ -97,7 +97,9 @@ int ttree_insert(struct ttree *tt, unsigned int *key, void *data)
 			}
 		}
 	}
-	(*c)->data = data;
+
+	if (*c)
+		(*c)->data = data;
 
 	return 1;
 }
Signature · 1 line
1.7.3.1
#25 lore raw

PATCH 12/30 Check for NULL pointer

In reply to Zdenek Kabelac

clangs is happier

---
 tools/reporter.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/reporter.c +1 -1
index f39d23f..eb0a857 100644
@@ -323,7 +323,7 @@ static int _report(struct cmd_context *cmd, int argc, char **argv,
 			log_error("Invalid options string: %s", opts);
 			return EINVALID_CMD_LINE;
 		}
-		if (*opts == '+') {
+		if (*opts == '+' && options) {
 			if (!(str = dm_pool_alloc(cmd->mem,
 					 strlen(options) + strlen(opts) + 1))) {
 				log_error("options string allocation failed");
Signature · 1 line
1.7.3.1
#26 lore raw

PATCH 13/30 Eliminate uninitialized_var

In reply to Zdenek Kabelac

Use some default value instead of a random number for unknown minor_version.

---
 lib/device/dev-md.c |    3 ++-
 1 files changed, 2 insertions(+), 1 deletions(-)
lib/device/dev-md.c +2 -1
index 89b9341..cb9feae 100644
@@ -59,7 +59,7 @@ typedef enum {
 
 static uint64_t _v1_sb_offset(uint64_t size, md_minor_version_t minor_version)
 {
-	uint64_t uninitialized_var(sb_offset);
+	uint64_t sb_offset;
 
 	switch(minor_version) {
 	case MD_MINOR_V0:
@@ -69,6 +69,7 @@ static uint64_t _v1_sb_offset(uint64_t size, md_minor_version_t minor_version)
 		sb_offset = 0;
 		break;
 	case MD_MINOR_V2:
+	default:
 		sb_offset = 4 * 2;
 		break;
 	}
Signature · 1 line
1.7.3.1
#30 lore raw

PATCH 14/30 Fix missing initilisation to 0

In reply to Zdenek Kabelac

Missing init value for 'found' which is later tested and may containe garbage value.

---
 daemons/cmirrord/cluster.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
daemons/cmirrord/cluster.c +1 -1
index dddaec9..b6c925a 100644
@@ -129,7 +129,7 @@ int cluster_send(struct clog_request *rq)
 {
 	int r;
 	int count=0;
-	int found;
+	int found = 0;
 	struct iovec iov;
 	struct clog_cpg *entry;
 
Signature · 1 line
1.7.3.1
#32 lore raw

PATCH 15/30 Function pull_state cannot work with NULL buffer

In reply to Zdenek Kabelac

In practice this situation should not happen in current lvm code.

---
 daemons/cmirrord/functions.c |    4 +++-
 1 files changed, 3 insertions(+), 1 deletions(-)
daemons/cmirrord/functions.c +3 -1
index e5da757..58e76ac 100644
@@ -1809,8 +1809,10 @@ int pull_state(const char *uuid, uint64_t luid,
 	int bitset_size;
 	struct log_c *lc;
 
-	if (!buf)
+	if (!buf) {
 		LOG_ERROR("pull_state: buf == NULL");
+		return -EINVAL;
+	}
 
 	lc = get_log(uuid, luid);
 	if (!lc) {
Signature · 1 line
1.7.3.1
#34 lore raw

PATCH 16/30 Fix void* pointer arithmetic

In reply to Zdenek Kabelac
---
 lib/mm/memlock.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
lib/mm/memlock.c +2 -2
index 3de8cc1..1e3065c 100644
@@ -87,8 +87,8 @@ static size_t _mstats; /* statistic for maps locking */
 static void _touch_memory(void *mem, size_t size)
 {
 	size_t pagesize = lvm_getpagesize();
-	void *pos = mem;
-	void *end = mem + size - sizeof(long);
+	char *pos = mem;
+	char *end = pos + size - sizeof(long);
 
 	while (pos < end) {
 		*(long *) pos = 1;
Signature · 1 line
1.7.3.1
#40 lore raw

PATCH 17/30 Use const pvid for device_from_pvid()

In reply to Zdenek Kabelac

Update interface for device_from_pvid and use const pvid.

---
 lib/cache/lvmcache.c |    2 +-
 lib/cache/lvmcache.h |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)
lib/cache/lvmcache.c +1 -1
index 35d6f76..83ac1a5 100644
@@ -733,7 +733,7 @@ struct dm_list *lvmcache_get_pvids(struct cmd_context *cmd, const char *vgname,
 	return pvids;
 }
 
-struct device *device_from_pvid(struct cmd_context *cmd, struct id *pvid,
+struct device *device_from_pvid(struct cmd_context *cmd, const struct id *pvid,
 				unsigned *scan_done_once)
 {
 	struct label *label;
lib/cache/lvmcache.h +1 -1
index 4966a5c..28f8541 100644
@@ -94,7 +94,7 @@ struct lvmcache_vginfo *vginfo_from_vgname(const char *vgname,
 struct lvmcache_vginfo *vginfo_from_vgid(const char *vgid);
 struct lvmcache_info *info_from_pvid(const char *pvid, int valid_only);
 const char *vgname_from_vgid(struct dm_pool *mem, const char *vgid);
-struct device *device_from_pvid(struct cmd_context *cmd, struct id *pvid,
+struct device *device_from_pvid(struct cmd_context *cmd, const struct id *pvid,
 				unsigned *scan_done_once);
 const char *pvid_from_devname(struct cmd_context *cmd,
 			      const char *dev_name);
Signature · 1 line
1.7.3.1
#42 lore raw

PATCH 18/30 Use const pointer for return value of dm_basename

In reply to Zdenek Kabelac

Fix pointer created from passed input const pointer.

---
 libdm/libdevmapper.h |    2 +-
 libdm/libdm-string.c |    6 +++---
 2 files changed, 4 insertions(+), 4 deletions(-)
libdm/libdevmapper.h +1 -1
index 7ed60be..e2f2673 100644
@@ -955,7 +955,7 @@ int dm_snprintf(char *buf, size_t bufsize, const char *format, ...)
 /*
  * Returns pointer to the last component of the path.
  */
-char *dm_basename(const char *path);
+const char *dm_basename(const char *path);
 
 /**************************
  * file/stream manipulation
libdm/libdm-string.c +3 -3
index ce05a10..d8fca1a 100644
@@ -123,11 +123,11 @@ int dm_snprintf(char *buf, size_t bufsize, const char *format, ...)
 	return n;
 }
 
-char *dm_basename(const char *path)
+const char *dm_basename(const char *path)
 {
-	char *p = strrchr(path, '/');
+	const char *p = strrchr(path, '/');
 
-	return p ? p + 1 : (char *) path;
+	return p ? p + 1 : path;
 }
 
 int dm_asprintf(char **result, const char *format, ...)
Signature · 1 line
1.7.3.1
#44 lore raw

PATCH 19/30 Fix constness warning

In reply to Zdenek Kabelac

After update of device_from_pvid API fix few constness warnings

---
 lib/cache/lvmcache.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)
lib/cache/lvmcache.c +4 -4
index 83ac1a5..60edcec 100644
@@ -514,7 +514,7 @@ char *lvmcache_vgname_from_pvid(struct cmd_context *cmd, const char *pvid)
 	struct lvmcache_info *info;
 	char *vgname;
 
-	if (!device_from_pvid(cmd, (struct id *)pvid, NULL)) {
+	if (!device_from_pvid(cmd, (const struct id *)pvid, NULL)) {
 		log_error("Couldn't find device with uuid %s.", pvid);
 		return NULL;
 	}
@@ -740,7 +740,7 @@ struct device *device_from_pvid(struct cmd_context *cmd, const struct id *pvid,
 	struct lvmcache_info *info;
 
 	/* Already cached ? */
-	if ((info = info_from_pvid((char *) pvid, 0))) {
+	if ((info = info_from_pvid((const char *) pvid, 0))) {
 		if (label_read(info->dev, &label, UINT64_C(0))) {
 			info = (struct lvmcache_info *) label->info;
 			if (id_equal(pvid, (struct id *) &info->dev->pvid))
@@ -751,7 +751,7 @@ struct device *device_from_pvid(struct cmd_context *cmd, const struct id *pvid,
 	lvmcache_label_scan(cmd, 0);
 
 	/* Try again */
-	if ((info = info_from_pvid((char *) pvid, 0))) {
+	if ((info = info_from_pvid((const char *) pvid, 0))) {
 		if (label_read(info->dev, &label, UINT64_C(0))) {
 			info = (struct lvmcache_info *) label->info;
 			if (id_equal(pvid, (struct id *) &info->dev->pvid))
@@ -767,7 +767,7 @@ struct device *device_from_pvid(struct cmd_context *cmd, const struct id *pvid,
 		*scan_done_once = 1;
 
 	/* Try again */
-	if ((info = info_from_pvid((char *) pvid, 0))) {
+	if ((info = info_from_pvid((const char *) pvid, 0))) {
 		if (label_read(info->dev, &label, UINT64_C(0))) {
 			info = (struct lvmcache_info *) label->info;
 			if (id_equal(pvid, (struct id *) &info->dev->pvid))
Signature · 1 line
1.7.3.1
#46 lore raw

PATCH 20/30 Fix constness warning

In reply to Zdenek Kabelac
---
 lib/metadata/metadata.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
lib/metadata/metadata.c +1 -1
index 2da4085..77fdd02 100644
@@ -3141,7 +3141,7 @@ struct logical_volume *lv_from_lvid(struct cmd_context *cmd, const char *lvid_s,
 	lvid = (const union lvid *) lvid_s;
 
 	log_very_verbose("Finding volume group for uuid %s", lvid_s);
-	if (!(vg = _vg_read_by_vgid(cmd, (char *)lvid->id[0].uuid, precommitted))) {
+	if (!(vg = _vg_read_by_vgid(cmd, (const char *)lvid->id[0].uuid, precommitted))) {
 		log_error("Volume group for uuid not found: %s", lvid_s);
 		return NULL;
 	}
Signature · 1 line
1.7.3.1
#48 lore raw

PATCH 21/30 Fix constness warning

In reply to Zdenek Kabelac
---
 lib/misc/crc.c |    6 +++---
 1 files changed, 3 insertions(+), 3 deletions(-)
lib/misc/crc.c +3 -3
index f2f2f06..9f95c37 100644
@@ -59,8 +59,8 @@ static uint32_t _calc_crc_new(uint32_t initial, const uint8_t *buf, uint32_t siz
 		0xbdbdf21c, 0xcabac28a, 0x53b39330, 0x24b4a3a6, 0xbad03605, 0xcdd70693, 0x54de5729, 0x23d967bf,
 		0xb3667a2e, 0xc4614ab8, 0x5d681b02, 0x2a6f2b94, 0xb40bbe37, 0xc30c8ea1, 0x5a05df1b, 0x2d02ef8d,
 	};
-	uint32_t *start = (uint32_t *) buf;
-	uint32_t *end = (uint32_t *) (buf + (size & 0xfffffffc));
+	const uint32_t *start = (const uint32_t *) buf;
+	const uint32_t *end = (const uint32_t *) (buf + (size & 0xfffffffc));
 	uint32_t crc = initial;
    
 	/* Process 4 bytes per iteration */
@@ -73,7 +73,7 @@ static uint32_t _calc_crc_new(uint32_t initial, const uint8_t *buf, uint32_t siz
 	}
 
 	/* Process any bytes left over */
-	buf = (uint8_t *) start;
+	buf = (const uint8_t *) start;
 	size = size & 0x3;
 	while (size--) {
 		crc = crc ^ *buf++;
Signature · 1 line
1.7.3.1
#50 lore raw

PATCH 22/30 Fix constness warning

In reply to Zdenek Kabelac

Fix usage of const 'data' pointer and also assign void* directly without uneeded cast for C.

---
 tools/dmsetup.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/dmsetup.c +1 -1
index a38fc4c..88291f7 100644
@@ -2399,7 +2399,7 @@ static int _dm_deps_disp(struct dm_report *rh, struct dm_pool *mem,
 			 struct dm_report_field *field, const void *data,
 			 void *private)
 {
-	struct dm_deps *deps = (struct dm_deps *) data;
+	const struct dm_deps *deps = data;
 	int i;
 	char buf[DM_MAX_TYPE_NAME], *repstr;
 
Signature · 1 line
1.7.3.1
#52 lore raw

PATCH 23/30 Using const config node

In reply to Zdenek Kabelac
---
 lib/striped/striped.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
lib/striped/striped.c +1 -1
index 01fbc1e..44f9774 100644
@@ -71,7 +71,7 @@ static int _striped_text_import_area_count(struct config_node *sn, uint32_t *are
 static int _striped_text_import(struct lv_segment *seg, const struct config_node *sn,
 			struct dm_hash_table *pv_hash)
 {
-	struct config_node *cn;
+	const struct config_node *cn;
 
 	if ((seg->area_count != 1) &&
 	    !get_config_uint32(sn, "stripe_size", &seg->stripe_size)) {
Signature · 1 line
1.7.3.1
#54 lore raw

PATCH 24/30 bufused is assigned 0 in preceding source line

In reply to Zdenek Kabelac

clang Idempotent operation

---
 lib/log/log.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
lib/log/log.c +1 -1
index 5c49586..86b4988 100644
@@ -359,7 +359,7 @@ void print_log(int level, const char *file, int line, int dm_errno,
 		_already_logging = 1;
 		memset(&buf, ' ', sizeof(buf));
 		bufused = 0;
-		if ((n = dm_snprintf(buf, sizeof(buf) - bufused - 1,
+		if ((n = dm_snprintf(buf, sizeof(buf) - 1,
 				      "%s:%d %s%s", file, line, log_command_name(),
 				      _msg_prefix)) == -1)
 			goto done;
Signature · 1 line
1.7.3.1
#56 lore raw

PATCH 25/30 Proper C declaration

In reply to Zdenek Kabelac
---
 daemons/dmeventd/dmeventd.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)
daemons/dmeventd/dmeventd.c +2 -2
index d179879..bbe6e10 100644
@@ -1477,7 +1477,7 @@ static void _process_request(struct dm_event_fifos *fifos)
 	dm_free(msg.data);
 }
 
-static void _process_initial_registrations()
+static void _process_initial_registrations(void)
 {
 	int i = 0;
 	char *reg;
@@ -1697,7 +1697,7 @@ static void _daemonize(void)
 	setsid();
 }
 
-static void restart()
+static void restart(void)
 {
 	struct dm_event_fifos fifos;
 	struct dm_event_daemon_message msg = { 0, 0, NULL };
Signature · 1 line
1.7.3.1
#58 lore raw

PATCH 26/30 Use static

In reply to Zdenek Kabelac
---
 libdm/mm/pool-fast.c |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)
libdm/mm/pool-fast.c +4 -4
index d9030cf..ebd982e 100644
@@ -33,8 +33,8 @@ struct dm_pool {
 	unsigned object_alignment;
 };
 
-void _align_chunk(struct chunk *c, unsigned alignment);
-struct chunk *_new_chunk(struct dm_pool *p, size_t s);
+static void _align_chunk(struct chunk *c, unsigned alignment);
+static struct chunk *_new_chunk(struct dm_pool *p, size_t s);
 static void _free_chunk(struct chunk *c);
 
 /* by default things come out aligned for doubles */
@@ -240,12 +240,12 @@ void dm_pool_abandon_object(struct dm_pool *p)
 	p->object_alignment = DEFAULT_ALIGNMENT;
 }
 
-void _align_chunk(struct chunk *c, unsigned alignment)
+static void _align_chunk(struct chunk *c, unsigned alignment)
 {
 	c->begin += alignment - ((unsigned long) c->begin & (alignment - 1));
 }
 
-struct chunk *_new_chunk(struct dm_pool *p, size_t s)
+static struct chunk *_new_chunk(struct dm_pool *p, size_t s)
 {
 	struct chunk *c;
 
Signature · 1 line
1.7.3.1
#60 lore raw

PATCH 27/30 Instrument compiler about code unreachability

In reply to Zdenek Kabelac

Clang needs some instrumentation for code analysis. It might probably help also for gcc.

---
 daemons/clvmd/clvmd.c |    7 +++++--
 1 files changed, 5 insertions(+), 2 deletions(-)
daemons/clvmd/clvmd.c +5 -2
index 57f7ffe..f9096cc 100644
@@ -491,6 +491,7 @@ int main(int argc, char *argv[])
 		DEBUGLOG("Can't initialise cluster interface\n");
 		log_error("Can't initialise cluster interface\n");
 		child_init_signal(DFAIL_CLUSTER_IF);
+		__builtin_unreachable();
 	}
 	DEBUGLOG("Cluster ready, doing some more initialisation\n");
 
@@ -505,8 +506,10 @@ int main(int argc, char *argv[])
 
 	/* Add the local socket to the list */
 	newfd = malloc(sizeof(struct local_client));
-	if (!newfd)
-	        child_init_signal(DFAIL_MALLOC);
+	if (!newfd) {
+		child_init_signal(DFAIL_MALLOC);
+		__builtin_unreachable();
+	}
 
 	newfd->fd = local_sock;
 	newfd->removeme = 0;
Signature · 1 line
1.7.3.1
#62 lore raw

PATCH 28/30 Skip uninitialized macro for scan-build

In reply to Zdenek Kabelac

scan-build reports assigment of ununitialised values.

---
 lib/misc/util.h |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)
lib/misc/util.h +4 -0
index fcda369..59e6980 100644
@@ -25,7 +25,11 @@
 		     (void) (&_a == &_b); \
 		     _a > _b ? _a : _b; })
 
+#ifndef  __clang__
 #define uninitialized_var(x) x = x
+#else
+#define uninitialized_var(x) x
+#endif
 
 #define KERNEL_VERSION(major, minor, release) (((major) << 16) + ((minor) << 8) + (release))
 
Signature · 1 line
1.7.3.1
#64 lore raw

PATCH 29/30 Add some __attribute__ instrumentation

In reply to Zdenek Kabelac

Add some malloc intrumentation and warning attributes.

---
 libdm/libdevmapper.h |   26 +++++++++++++-------------
 1 files changed, 13 insertions(+), 13 deletions(-)
libdm/libdevmapper.h +13 -13
index e2f2673..9ae8bbd 100644
@@ -510,13 +510,13 @@ uint32_t dm_tree_get_cookie(struct dm_tree_node *node);
  * Memory management
  *******************/
 
-void *dm_malloc_aux(size_t s, const char *file, int line);
-void *dm_malloc_aux_debug(size_t s, const char *file, int line);
-void *dm_zalloc_aux(size_t s, const char *file, int line);
-void *dm_zalloc_aux_debug(size_t s, const char *file, int line);
-char *dm_strdup_aux(const char *str, const char *file, int line);
+void *dm_malloc_aux(size_t s, const char *file, int line) __attribute__((malloc)) __attribute__((__warn_unused_result__));
+void *dm_malloc_aux_debug(size_t s, const char *file, int line) __attribute__((__warn_unused_result__));
+void *dm_zalloc_aux(size_t s, const char *file, int line)  __attribute__((malloc)) __attribute__((__warn_unused_result__));
+void *dm_zalloc_aux_debug(size_t s, const char *file, int line) __attribute__((__warn_unused_result__));
+char *dm_strdup_aux(const char *str, const char *file, int line) __attribute__((malloc)) __attribute__((__warn_unused_result__));
 void dm_free_aux(void *p);
-void *dm_realloc_aux(void *p, unsigned int s, const char *file, int line);
+void *dm_realloc_aux(void *p, unsigned int s, const char *file, int line) __attribute__((__warn_unused_result__));
 int dm_dump_memory_debug(void);
 void dm_bounds_check_debug(void);
 
@@ -579,12 +579,12 @@ void dm_bounds_check_debug(void);
 struct dm_pool;
 
 /* constructor and destructor */
-struct dm_pool *dm_pool_create(const char *name, size_t chunk_hint);
+struct dm_pool *dm_pool_create(const char *name, size_t chunk_hint) __attribute__((__warn_unused_result__));
 void dm_pool_destroy(struct dm_pool *p);
 
 /* simple allocation/free routines */
-void *dm_pool_alloc(struct dm_pool *p, size_t s);
-void *dm_pool_alloc_aligned(struct dm_pool *p, size_t s, unsigned alignment);
+void *dm_pool_alloc(struct dm_pool *p, size_t s) __attribute__((__warn_unused_result__));
+void *dm_pool_alloc_aligned(struct dm_pool *p, size_t s, unsigned alignment) __attribute__((__warn_unused_result__));
 void dm_pool_empty(struct dm_pool *p);
 void dm_pool_free(struct dm_pool *p, void *ptr);
 
@@ -639,9 +639,9 @@ void *dm_pool_end_object(struct dm_pool *p);
 void dm_pool_abandon_object(struct dm_pool *p);
 
 /* utilities */
-char *dm_pool_strdup(struct dm_pool *p, const char *str);
-char *dm_pool_strndup(struct dm_pool *p, const char *str, size_t n);
-void *dm_pool_zalloc(struct dm_pool *p, size_t s);
+char *dm_pool_strdup(struct dm_pool *p, const char *str) __attribute__((__warn_unused_result__));
+char *dm_pool_strndup(struct dm_pool *p, const char *str, size_t n) __attribute__((__warn_unused_result__));
+void *dm_pool_zalloc(struct dm_pool *p, size_t s) __attribute__((__warn_unused_result__));
 
 /******************
  * bitset functions
@@ -699,7 +699,7 @@ struct dm_hash_node;
 
 typedef void (*dm_hash_iterate_fn) (void *data);
 
-struct dm_hash_table *dm_hash_create(unsigned size_hint);
+struct dm_hash_table *dm_hash_create(unsigned size_hint) __attribute__((__warn_unused_result__));
 void dm_hash_destroy(struct dm_hash_table *t);
 void dm_hash_wipe(struct dm_hash_table *t);
 
Signature · 1 line
1.7.3.1
#67 lore raw

PATCH 30/30 __attribute__((nonnull(1)))

In reply to Zdenek Kabelac
---
 lib/device/dev-cache.h |    2 +-
 libdm/regex/ttree.c    |    3 ++-
 2 files changed, 3 insertions(+), 2 deletions(-)
lib/device/dev-cache.h +1 -1
index c1c86d6..772d28e 100644
@@ -41,7 +41,7 @@ int dev_cache_has_scanned(void);
 
 int dev_cache_add_dir(const char *path);
 int dev_cache_add_loopfile(const char *path);
-struct device *dev_cache_get(const char *name, struct dev_filter *f);
+struct device *dev_cache_get(const char *name, struct dev_filter *f) __attribute__((nonnull(1)));
 
 void dev_set_preferred_name(struct str_list *sl, struct device *dev);
 
libdm/regex/ttree.c +2 -1
index 0ad40bd..fc24c96 100644
@@ -28,7 +28,8 @@ struct ttree {
 	struct node *root;
 };
 
-static struct node **_lookup_single(struct node **c, unsigned int k)
+static __attribute__((nonnull (1)))
+struct node **_lookup_single(struct node **c, unsigned int k)
 {
 	while (*c) {
 		if (k < (*c)->k)
Signature · 1 line
1.7.3.1
#69 lore raw

cover Fixes from clang

In reply to Zdenek Kabelac
Quoted 37 lines · Zdenek Kabelac

Patchset combines code cleanups for compiler warnings and code fixes found by static analyzer scan-build.

Potentional NULL pointer dereferencies are mostly in very unusual code execution paths. I've tried to order them on their importance - the later ones are not event trigerable as surrounding code should not be to pass problematic values - but checks seems to be simple enough to keep static analysis happy.

Last patches in this set add some instrumentation to function calls that might allow better code optimization or better static analysis.

Zdenek Kabelac (30):
Fix clang warning for ntohl(*((uint32_t *)buf)) Use dm_pool_grow_object embeded strlen for 0 delta Check (type) is not NULL before access
Fix NULL pointer dereference
Print vg_name and do not to access vg->name Ensure we always have origin defined Ensure seg is nonnull
Ensure first is not NULL
Reuse result of previous strchr.
Fix potential NULL pointer dereference
Fix theoretical usage of NULL pointer dereference Check for NULL pointer
Eliminate uninitialized_var
Fix missing initilisation to 0
Function pull_state cannot work with NULL buffer Fix void* pointer arithmetic
Use const pvid for device_from_pvid()
Use const pointer for return value of dm_basename Fix constness warning
Fix constness warning
Fix constness warning
Fix constness warning

Quoted 8 lines

Using const config node
bufused is assigned 0 in preceding source line Proper C declaration
Use static
Instrument compiler about code unreachability Skip uninitialized macro for scan-build
Add some __attribute__ instrumentation __attribute__((nonnull(1)))

Nice work!
Getting any project in good enough shape to pass clang's tests is tedious but well worth it.

#2 lore raw

PATCH 1/30 Fix clang warning for ntohl(*((uint32_t *)buf))

In reply to Zdenek Kabelac

At Mon, 25 Oct 2010 10:24:08 +0200,

Quoted 9 lines · Zdenek Kabelac

We cast (char*) to (uint32_t*) that changes alignment requierements. For our case the code has been correct as alloca() returns properly aligned buffer, however this patch make it cleaner and more readable and avoids warning generation.

FIXME: code duplication

Ack

#4 lore raw

PATCH 2/30 Use dm_pool_grow_object embeded strlen for 0 delta

In reply to Zdenek Kabelac

At Mon, 25 Oct 2010 10:24:09 +0200,

Quoted 4 lines · Zdenek Kabelac

Use internal strlen for 0

Ack, but I don't think this improves the readablility of the code. The problem is the extra semantics of zero as the third arg to dm_pool_grow_object() (probably my bad). Maybe introduce a new pool method dm_pool_grow_string(struct dm_pool *p, const char *str) ?

#8 lore raw

PATCH 3/30 Check (type) is not NULL before access

In reply to Zdenek Kabelac

At Mon, 25 Oct 2010 10:24:10 +0200,

Quoted 2 lines · Zdenek Kabelac

clang Logic error Dereference of null pointer

Should a generic warning be given if |type| is NULL?

#16 lore raw

PATCH 7/30 Ensure seg is nonnull

In reply to Zdenek Kabelac

At Mon, 25 Oct 2010 10:24:14 +0200,

Quoted 5 lines · Zdenek Kabelac

clang Logic error Dereference of null pointer
Make sure we do not try to use NULL seg pointer.

Does |seg| being NULL indicate a more serious problem? Should we fail rather than making the best of it and carrying on silently?

#68 lore raw

PATCH 30/30 __attribute__((nonnull(1)))

In reply to Zdenek Kabelac

At Mon, 25 Oct 2010 10:24:37 +0200,

Quoted 26 lines · Zdenek Kabelac
---
 lib/device/dev-cache.h |    2 +-
 libdm/regex/ttree.c    |    3 ++-
 2 files changed, 3 insertions(+), 2 deletions(-)
lib/device/dev-cache.h +1 -1
index c1c86d6..772d28e 100644
@@ -41,7 +41,7 @@ int dev_cache_has_scanned(void);
 
 int dev_cache_add_dir(const char *path);
 int dev_cache_add_loopfile(const char *path);
-struct device *dev_cache_get(const char *name, struct dev_filter *f);
+struct device *dev_cache_get(const char *name, struct dev_filter *f) __attribute__((nonnull(1)));
 
 void dev_set_preferred_name(struct str_list *sl, struct device *dev);
 
libdm/regex/ttree.c +0 -0
index 0ad40bd..fc24c96 100644

libdm/regex/ttree.c struct ttree { · line 32

Quoted 6 lines
libdm/regex/ttree.c +2 -1
@@ -28,7 +28,8 @@ struct ttree {
 	struct node *root;
 };
 
-static struct node **_lookup_single(struct node **c, unsigned int k)
+static __attribute__((nonnull (1)))
+struct node **_lookup_single(struct node **c, unsigned int k)

Ack. Verbose, isn't it.

#5 lore raw

PATCH 2/30 Use dm_pool_grow_object embeded strlen for 0 delta

In reply to [email protected]

Dne 25.10.2010 11:56, ejt at redhat.com napsal(a):

Quoted 8 lines

At Mon, 25 Oct 2010 10:24:09 +0200,

Quoted 4 lines · Zdenek Kabelac

Use internal strlen for 0

Ack, but I don't think this improves the readablility of the code. The problem is the extra semantics of zero as the third arg to dm_pool_grow_object() (probably my bad). Maybe introduce a new pool method dm_pool_grow_string(struct dm_pool *p, const char *str) ?

Well you are not the first one with this idea :)

Me and Milan thinks in the same way - if everyone agree - I'll make another version with ...grow_string()

Zdenek

#28 lore raw

PATCH 13/30 Eliminate uninitialized_var

In reply to [email protected]

Dne 25.10.2010 12:18, ejt at redhat.com napsal(a):

Quoted 9 lines

At Mon, 25 Oct 2010 10:24:20 +0200,

Quoted 5 lines · Zdenek Kabelac

Use some default value instead of a random number for unknown minor_version.

Nack. What if your default |sb_offset| is wrong?

Still better something defined - then completely random value.

But this patch is rather about make compiler analyzer quiet - maybe default: assert() for 'generic' version ?

Quoted 1 line

From the source code - we are actually using only MD_MINOR_V0,

as is: minor = MD_MINOR_VERSION_MIN (== MD_MINOR_V0)

Zdenek

#36 lore raw

PATCH 16/30 Fix void* pointer arithmetic

In reply to [email protected]

Dne 25.10.2010 12:20, ejt at redhat.com napsal(a):

Quoted 7 lines

At Mon, 25 Oct 2010 10:24:23 +0200,

Quoted 3 lines · Zdenek Kabelac

Nack. No explanation. eg, Are we deprecating the void pointer arithmetic gcc extension?

void* arithmetic is IMHO bad practice and we should avoid it's usage. char* clearly defines 1byte - I'd prefer to use standard C whenever we can. (And gcc can show this as warning - so why not fixing it)

Zdenek

#29 lore raw

PATCH 13/30 Eliminate uninitialized_var

In reply to Zdenek Kabelac

At Mon, 25 Oct 2010 13:13:13 +0200,

Quoted 2 lines · Zdenek Kabelac

But this patch is rather about make compiler analyzer quiet - maybe default: assert() for 'generic' version ?

Or just plain abort() so it doesn't get compiled out in the release version?

#37 lore raw

PATCH 16/30 Fix void* pointer arithmetic

In reply to Zdenek Kabelac

At Mon, 25 Oct 2010 13:17:05 +0200,

Quoted 6 lines · Zdenek Kabelac

Dne 25.10.2010 12:20, ejt at redhat.com napsal(a):

Quoted 4 lines

At Mon, 25 Oct 2010 10:24:23 +0200,

Quoted 2 lines · Zdenek Kabelac
Quoted 7 lines
Quoted 3 lines
Quoted 1 line

Nack. No explanation. eg, Are we deprecating the void pointer arithmetic gcc extension?

void* arithmetic is IMHO bad practice and we should avoid it's usage. char* clearly defines 1byte - I'd prefer to use standard C whenever we can. (And gcc can show this as warning - so why not fixing it)

Ack

#38 lore raw

PATCH 16/30 Fix void* pointer arithmetic

In reply to Zdenek Kabelac

lib/mm/memlock.c static size_t _mstats; /* statistic for maps locking */ · line 91

Quoted 7 lines · Zdenek Kabelac

static void _touch_memory(void *mem, size_t size) {

lib/mm/memlock.c +2 -2
 	size_t pagesize = lvm_getpagesize();
-	void *pos = mem;
-	void *end = mem + size - sizeof(long);
+	char *pos = mem;
+	char *end = pos + size - sizeof(long);

And what about trash this "touch memory" code completely?

The whole idea of touching memory before locking is IMHO obsolete.

Locking memory should do this implicitly - can we prove that this code has some effect on recent systems?

Milan

#39 lore raw

PATCH 16/30 Fix void* pointer arithmetic

Dne 25.10.2010 13:23, Milan Broz napsal(a):

Quoted 12 lines · Zdenek Kabelac

lib/mm/memlock.c static size_t _mstats; /* statistic for maps locking */ · line 91

Quoted 8 lines

static void _touch_memory(void *mem, size_t size) {

lib/mm/memlock.c +2 -2
 	size_t pagesize = lvm_getpagesize();
-	void *pos = mem;
-	void *end = mem + size - sizeof(long);
+	char *pos = mem;
+	char *end = pos + size - sizeof(long);

And what about trash this "touch memory" code completely?

Quoted 5 lines

The whole idea of touching memory before locking is IMHO obsolete.

Locking memory should do this implicitly - can we prove that this code has some effect on recent systems?

Our memlock code should be physically allocating and paging all the memory to the process space - so the only difference is IMHO at which point in time memory failure happens - with memory touching - it might occur sooner in some cases - but we lose some speed by trashing CPU cache and flagging memory pages.

Zdenek

#66 lore raw

PATCH 29/30 Add some __attribute__ instrumentation

In reply to [email protected]

Dne 25.10.2010 12:44, ejt at redhat.com napsal(a):

Quoted 9 lines

At Mon, 25 Oct 2010 10:24:36 +0200,

Quoted 4 lines · Zdenek Kabelac

Add some malloc intrumentation and warning attributes.

Ack. But I question how far you intend to go with this. Surely it's more unusual to return a value that can be ignored?

I think wherever it is an error to ignore returned value on public API function, we should put some warning. In this case it must lead to memory leak - so it's good idea to give the user this knowledge in compile time.

(btw I'm doing nothing more then what original system headers do for you :))

What seems to have some 'hidden' fruits are those 'nonnull' __attributes__. We have quite a few functions which take a pointer and expect it to be non-null - so we could mark these functions and get static analysis to be done on them.

Zdenek