Skip to contentSkip to conversation

nvme-fabrics: small cleanup

PATCH Applied Chaitanya Kulkarni 5 participants 16 messages Jan 12, 2022 – Jan 26, 2022 NVMe

lore
Share & tools
lore
#0 lore raw

cover nvme-fabrics: small cleanup

From: Chaitanya Kulkarni <[email protected]>

Hi,

This has small cleaups and a fix suggested by Sagi for nvmf_dev_show().

-ck

Chaitanya Kulkarni (5):
nvme-fabrics: use consistent zeroout pattern
nvme-fabrics: remove unnecessary braces for case
nvme-fabrics: use unsigned int type
nvme-fabrics: use unsigned int type
nvme-fabrics: set ret to -ENODEV for error case

 drivers/nvme/host/fabrics.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
Signature · 1 line
2.29.0
#1 lore raw

PATCH 1/5 nvme-fabrics: use consistent zeroout pattern

From: Chaitanya Kulkarni <[email protected]>

Remove zeroout memeset call & zeroout local variable cmd at the time of declaration in nvmf_ref_read32() similar to what we have done in nvmf_reg_read64(), nvmf_reg_write32(), nvmf_connect_admin_queue(), and nvmf_connect_io_queue().

---
 drivers/nvme/host/fabrics.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
drivers/nvme/host/fabrics.c +1 -2
index 7ae041e2b3fb..a0640986add9 100644
@@ -144,11 +144,10 @@ EXPORT_SYMBOL_GPL(nvmf_get_address);
  */
 int nvmf_reg_read32(struct nvme_ctrl *ctrl, u32 off, u32 *val)
 {
-	struct nvme_command cmd;
+	struct nvme_command cmd = { };
 	union nvme_result res;
 	int ret;
 
-	memset(&cmd, 0, sizeof(cmd));
 	cmd.prop_get.opcode = nvme_fabrics_command;
 	cmd.prop_get.fctype = nvme_fabrics_type_property_get;
 	cmd.prop_get.offset = cpu_to_le32(off);
Signature · 1 line
2.29.0
maintainer Reviewed-by on v1 1/5
#2 lore raw

Re: [PATCH 1/5] nvme-fabrics: use consistent zeroout pattern

Quoted 8 lines · Chaitanya Kulkarni

From: Chaitanya Kulkarni <[email protected]>

Remove zeroout memeset call & zeroout local variable cmd at the time of declaration in nvmf_ref_read32() similar to what we have done in nvmf_reg_read64(), nvmf_reg_write32(), nvmf_connect_admin_queue(), and nvmf_connect_io_queue().

Looks good.

#3 lore raw

PATCH 2/5 nvme-fabrics: remove unnecessary braces for case

In reply to Chaitanya Kulkarni

From: Chaitanya Kulkarni <[email protected]>

Braces are not required for enum value NVME_SC_CONNECT_INVALID_PARAM when used on the switch-case statement, remove the braces.

---
 drivers/nvme/host/fabrics.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/nvme/host/fabrics.c +1 -1
index a0640986add9..d38a630ce3dc 100644
@@ -271,7 +271,7 @@ static void nvmf_log_connect_error(struct nvme_ctrl *ctrl,
 	int err_sctype = errval & ~NVME_SC_DNR;
 
 	switch (err_sctype) {
-	case (NVME_SC_CONNECT_INVALID_PARAM):
+	case NVME_SC_CONNECT_INVALID_PARAM:
 		if (offset >> 16) {
 			char *inv_data = "Connect Invalid Data Parameter";
 
Signature · 1 line
2.29.0
#5 lore raw

PATCH 3/5 nvme-fabrics: use unsigned int type

In reply to Chaitanya Kulkarni

From: Chaitanya Kulkarni <[email protected]>

Loop variable i will never have a negative value, so use unsigned int type instaed of int.

---
 drivers/nvme/host/fabrics.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/nvme/host/fabrics.c +1 -1
index d38a630ce3dc..90b3474ea276 100644
@@ -872,7 +872,7 @@ static int nvmf_check_required_opts(struct nvmf_ctrl_options *opts,
 		unsigned int required_opts)
 {
 	if ((opts->mask & required_opts) != required_opts) {
-		int i;
+		unsigned int i;
 
 		for (i = 0; i < ARRAY_SIZE(opt_tokens); i++) {
 			if ((opt_tokens[i].token & required_opts) &&
Signature · 1 line
2.29.0
#6 lore raw

Re: [PATCH 3/5] nvme-fabrics: use unsigned int type

Quoted 15 lines · Chaitanya Kulkarni

From: Chaitanya Kulkarni <[email protected]>

Loop variable i will never have a negative value, so use unsigned int type instaed of int.

---
  drivers/nvme/host/fabrics.c | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)
drivers/nvme/host/fabrics.c +0 -0
index d38a630ce3dc..90b3474ea276 100644

drivers/nvme/host/fabrics.c static int nvmf_check_required_opts(struct nvmf_ctrl_options *opts, · line 878

Quoted 8 lines
drivers/nvme/host/fabrics.c +1 -1
@@ -872,7 +872,7 @@ static int nvmf_check_required_opts(struct nvmf_ctrl_options *opts,
  		unsigned int required_opts)
  {
  	if ((opts->mask & required_opts) != required_opts) {
-		int i;
+		unsigned int i;
 
  		for (i = 0; i < ARRAY_SIZE(opt_tokens); i++) {
  			if ((opt_tokens[i].token & required_opts) &&

I'm not against this but this means that we should replace the entire kernel that uses int inside a for loop.

If we decide to go with this approach - patch 3/5 and 4/5 should be squashed.

maintainer
#7 lore raw

Re: [PATCH 3/5] nvme-fabrics: use unsigned int type

In reply to Chaitanya Kulkarni
Quoted 6 lines · Chaitanya Kulkarni

From: Chaitanya Kulkarni <[email protected]>

Loop variable i will never have a negative value, so use unsigned int type instaed of int.

Just IMO, I do not particularly care for this kind of change. The driver and kernel use 'int i' for loop control in so many places. Granted, 'unsigned i' is also used too...

#8 lore raw

PATCH 4/5 nvme-fabrics: use unsigned int type

In reply to Chaitanya Kulkarni

From: Chaitanya Kulkarni <[email protected]>

Loop variable i will never have a negative value, so use unsigned int type instaed of int.

---
 drivers/nvme/host/fabrics.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/nvme/host/fabrics.c +1 -1
index 90b3474ea276..df0326db306f 100644
@@ -922,7 +922,7 @@ static int nvmf_check_allowed_opts(struct nvmf_ctrl_options *opts,
 		unsigned int allowed_opts)
 {
 	if (opts->mask & ~allowed_opts) {
-		int i;
+		unsigned int i;
 
 		for (i = 0; i < ARRAY_SIZE(opt_tokens); i++) {
 			if ((opt_tokens[i].token & opts->mask) &&
Signature · 1 line
2.29.0
#9 lore raw

PATCH 5/5 nvme-fabrics: set ret to -ENODEV for error case

In reply to Chaitanya Kulkarni

From: Chaitanya Kulkarni <[email protected]>

In function nvmf_dev_show() if we don't find the controller set the ret variable to -ENODEV.

---
 drivers/nvme/host/fabrics.c | 1 +
 1 file changed, 1 insertion(+)
drivers/nvme/host/fabrics.c +1 -0
index df0326db306f..ca23ac926710 100644
@@ -1097,6 +1097,7 @@ static int nvmf_dev_show(struct seq_file *seq_file, void *private)
 	ctrl = seq_file->private;
 	if (!ctrl) {
 		__nvmf_concat_opt_tokens(seq_file);
+		ret = -ENODEV;
 		goto out_unlock;
 	}
 
Signature · 1 line
2.29.0
maintainer
#10 lore raw

Re: [PATCH 5/5] nvme-fabrics: set ret to -ENODEV for error case

Quoted 16 lines · Chaitanya Kulkarni

From: Chaitanya Kulkarni <[email protected]>

In function nvmf_dev_show() if we don't find the controller set the ret variable to -ENODEV.

---
 drivers/nvme/host/fabrics.c | 1 +
 1 file changed, 1 insertion(+)
drivers/nvme/host/fabrics.c +0 -0
index df0326db306f..ca23ac926710 100644

drivers/nvme/host/fabrics.c static int nvmf_dev_show(struct seq_file *seq_file, void *private) · line 1102

Quoted 6 lines
drivers/nvme/host/fabrics.c +1 -0
@@ -1097,6 +1097,7 @@ static int nvmf_dev_show(struct seq_file *seq_file, void *private)
 	ctrl = seq_file->private;
 	if (!ctrl) {
 		__nvmf_concat_opt_tokens(seq_file);
+		ret = -ENODEV;
 		goto out_unlock;
 	}

I missed the discussion on this. Doesn't the __nvmf_concat_opt_tokens() already return parameters that indicate to the caller that no controller was found?

#11 lore raw

Re: [PATCH 5/5] nvme-fabrics: set ret to -ENODEV for error case

Keith,

Quoted 15 lines · Keith Busch

External email: Use caution opening links or attachments

Quoted 11 lines · Chaitanya Kulkarni

From: Chaitanya Kulkarni <[email protected]>

In function nvmf_dev_show() if we don't find the controller set the ret variable to -ENODEV.

---
  drivers/nvme/host/fabrics.c | 1 +
  1 file changed, 1 insertion(+)
Quoted 16 lines

drivers/nvme/host/fabrics.c static int nvmf_dev_show(struct seq_file *seq_file, void *private) · line 1102

Quoted 11 lines
drivers/nvme/host/fabrics.c +1 -0
index df0326db306f..ca23ac926710 100644
@@ -1097,6 +1097,7 @@ static int nvmf_dev_show(struct seq_file *seq_file, void *private)
       ctrl = seq_file->private;
       if (!ctrl) {
               __nvmf_concat_opt_tokens(seq_file);
+             ret = -ENODEV;
               goto out_unlock;
       }

I missed the discussion on this. Doesn't the __nvmf_concat_opt_tokens() already return parameters that indicate to the caller that no controller was found?

The function I found in the repo returns nothing, is that the one you are referring to ?

static void __nvmf_concat_opt_tokens(struct seq_file *seq_file)

-ck

maintainer
#12 lore raw

Re: [PATCH 5/5] nvme-fabrics: set ret to -ENODEV for error case

Quoted 29 lines
Quoted 12 lines
Quoted 12 lines

From: Chaitanya Kulkarni <[email protected]>

In function nvmf_dev_show() if we don't find the controller set the ret variable to -ENODEV.

---
   drivers/nvme/host/fabrics.c | 1 +
   1 file changed, 1 insertion(+)
Quoted 16 lines

drivers/nvme/host/fabrics.c static int nvmf_dev_show(struct seq_file *seq_file, void *private) · line 1102

Quoted 11 lines
drivers/nvme/host/fabrics.c +1 -0
index df0326db306f..ca23ac926710 100644
@@ -1097,6 +1097,7 @@ static int nvmf_dev_show(struct seq_file *seq_file, void *private)
        ctrl = seq_file->private;
        if (!ctrl) {
                __nvmf_concat_opt_tokens(seq_file);
+             ret = -ENODEV;
                goto out_unlock;
        }

I missed the discussion on this. Doesn't the __nvmf_concat_opt_tokens() already return parameters that indicate to the caller that no controller was found?

Quoted 4 lines

The function I found in the repo returns nothing, is that the one you are referring to ?

static void __nvmf_concat_opt_tokens(struct seq_file *seq_file)

This is the path that Hannes added for userspace to detect kernel capabilities. So probably its not appropriate to return an error here, but rather take care of the original warning instead.

#13 lore raw

Re: [PATCH 5/5] nvme-fabrics: set ret to -ENODEV for error case

Quoted 6 lines
Quoted 6 lines
Quoted 5 lines

I missed the discussion on this. Doesn't the __nvmf_concat_opt_tokens() already return parameters that indicate to the caller that no controller was found?

Quoted 8 lines
Quoted 4 lines

The function I found in the repo returns nothing, is that the one you are referring to ?

static void __nvmf_concat_opt_tokens(struct seq_file *seq_file)

This is the path that Hannes added for userspace to detect kernel capabilities. So probably its not appropriate to return an error here, but rather take care of the original warning instead.

When I look at the code it looks unconventional to not return error when controller not found case, why is it not appropriate to return error can you please explain ?

Can we add a meaningful comment to avoid future such patches.

maintainer
#14 lore raw

Re: [PATCH 5/5] nvme-fabrics: set ret to -ENODEV for error case

Quoted 15 lines
Quoted 6 lines
Quoted 6 lines
Quoted 5 lines

I missed the discussion on this. Doesn't the __nvmf_concat_opt_tokens() already return parameters that indicate to the caller that no controller was found?

Quoted 8 lines
Quoted 4 lines

The function I found in the repo returns nothing, is that the one you are referring to ?

static void __nvmf_concat_opt_tokens(struct seq_file *seq_file)

This is the path that Hannes added for userspace to detect kernel capabilities. So probably its not appropriate to return an error here, but rather take care of the original warning instead.

Quoted 5 lines

When I look at the code it looks unconventional to not return error when controller not found case, why is it not appropriate to return error can you please explain ?

Can we add a meaningful comment to avoid future such patches.

That is how Hannes proposed to read capabilities from the kernel in terms of valid connection string arguments. Read /dev/nvme-fabrics without a controller.