Skip to contentSkip to conversation

ipmi: make static checkers happy about range checks

PATCH In review Dan Carpenter 2 participants 2 messages Sep 17, 2010 – Sep 18, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ipmi: make static checkers happy about range checks

I'm working on a Smatch check that complains that addr->channel comes from the user and it might be negative. It actually can *not* be negative because we already checked that when we called ipmi_validate_addr() from handle_send_req(). Also it can't be greater than or equal to IPMI_MAX_CHANNELS because we already verified that as well.

Although this change doesn't do anything, it's still a cleanup because it makes the checking consistent and also it silences the static analysis warnings.

---
Really, I would prefer to just make ->channel unsigned but the type comes from the IPMI spec.

drivers/char/ipmi/ipmi_msghandler.c +3 -3
index 4f3f8c9..57f4c80 100644
@@ -1527,7 +1527,7 @@ static int i_ipmi_request(ipmi_user_t          user,
 		long                  seqid;
 		int                   broadcast = 0;
 
-		if (addr->channel >= IPMI_MAX_CHANNELS) {
+		if (addr->channel < 0 || addr->channel >= IPMI_MAX_CHANNELS) {
 			ipmi_inc_stat(intf, sent_invalid_commands);
 			rv = -EINVAL;
 			goto out_err;
@@ -1657,7 +1657,7 @@ static int i_ipmi_request(ipmi_user_t          user,
 		unsigned char         ipmb_seq;
 		long                  seqid;
 
-		if (addr->channel >= IPMI_MAX_CHANNELS) {
+		if (addr->channel < 0 || addr->channel >= IPMI_MAX_CHANNELS) {
 			ipmi_inc_stat(intf, sent_invalid_commands);
 			rv = -EINVAL;
 			goto out_err;
@@ -1797,7 +1797,7 @@ static int check_addr(ipmi_smi_t       intf,
 		      unsigned char    *saddr,
 		      unsigned char    *lun)
 {
-	if (addr->channel >= IPMI_MAX_CHANNELS)
+	if (addr->channel < 0 || addr->channel >= IPMI_MAX_CHANNELS)
 		return -EINVAL;
 	*lun = intf->channels[addr->channel].lun;
 	*saddr = intf->channels[addr->channel].address;
#1 lore raw

I'm ok with this, I don't think it will make any real difference besides the compiler converting that comparison to unsigned. I wish I had made that unsigned.

-corey

Quoted 48 lines · Dan Carpenter

I'm working on a Smatch check that complains that addr->channel comes from the user and it might be negative. It actually can *not* be negative because we already checked that when we called ipmi_validate_addr() from handle_send_req(). Also it can't be greater than or equal to IPMI_MAX_CHANNELS because we already verified that as well.

Although this change doesn't do anything, it's still a cleanup because it makes the checking consistent and also it silences the static analysis warnings.

---
Really, I would prefer to just make ->channel unsigned but the type comes from the IPMI spec.

drivers/char/ipmi/ipmi_msghandler.c +3 -3
index 4f3f8c9..57f4c80 100644
@@ -1527,7 +1527,7 @@ static int i_ipmi_request(ipmi_user_t          user,
  		long                  seqid;
  		int                   broadcast = 0;
 
-		if (addr->channel>= IPMI_MAX_CHANNELS) {
+		if (addr->channel<  0 || addr->channel>= IPMI_MAX_CHANNELS) {
  			ipmi_inc_stat(intf, sent_invalid_commands);
  			rv = -EINVAL;
  			goto out_err;
@@ -1657,7 +1657,7 @@ static int i_ipmi_request(ipmi_user_t          user,
  		unsigned char         ipmb_seq;
  		long                  seqid;
 
-		if (addr->channel>= IPMI_MAX_CHANNELS) {
+		if (addr->channel<  0 || addr->channel>= IPMI_MAX_CHANNELS) {
  			ipmi_inc_stat(intf, sent_invalid_commands);
  			rv = -EINVAL;
  			goto out_err;
@@ -1797,7 +1797,7 @@ static int check_addr(ipmi_smi_t       intf,
  		      unsigned char    *saddr,
  		      unsigned char    *lun)
  {
-	if (addr->channel>= IPMI_MAX_CHANNELS)
+	if (addr->channel<  0 || addr->channel>= IPMI_MAX_CHANNELS)
  		return -EINVAL;
  	*lun = intf->channels[addr->channel].lun;
  	*saddr = intf->channels[addr->channel].address;