[PATCH] btrfs-progs: reserve < 0 return value for errnos

[PATCH] btrfs-progs: reserve < 0 return value for errnos

[Anand Jain]

btrfs_read_dev_super() returns -1 upon not finding a suitable SB,
change that to return 1 instead, so that it can reserve the < 0
values for the errno communications.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
---
An independent patch, not related to any recent patch sent to ML.

 disk-io.c | 2 +-
 utils.c   | 2 +-
 volumes.c | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/disk-io.c b/disk-io.c
index f5edc4796619..ba87bb4ce6da 100644
--- a/disk-io.c
+++ b/disk-io.c
@@ -1491,7 +1491,7 @@ int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
 		}
 	}
 
-	return transid > 0 ? 0 : -1;
+	return transid > 0 ? 0 : 1;
 }
 
 static int write_dev_supers(struct btrfs_fs_info *fs_info,
diff --git a/utils.c b/utils.c
index 524f463d3140..7574cda8151c 100644
--- a/utils.c
+++ b/utils.c
@@ -1678,7 +1678,7 @@ int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
 		disk_super = (struct btrfs_super_block *)buf;
 		ret = btrfs_read_dev_super(fd, disk_super,
 					   BTRFS_SUPER_INFO_OFFSET, 0);
-		if (ret < 0) {
+		if (ret) {
 			ret = -EIO;
 			goto out;
 		}
diff --git a/volumes.c b/volumes.c
index 2209e5a9100b..87008bfdd586 100644
--- a/volumes.c
+++ b/volumes.c
@@ -268,7 +268,7 @@ int btrfs_scan_one_device(int fd, const char *path,
 
 	disk_super = (struct btrfs_super_block *)buf;
 	ret = btrfs_read_dev_super(fd, disk_super, super_offset, sbflags);
-	if (ret < 0)
+	if (ret)
 		return -EIO;
 	devid = btrfs_stack_device_id(&disk_super->dev_item);
 	if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_METADUMP)
-- 
2.13.1

Re: [PATCH] btrfs-progs: reserve < 0 return value for errnos

[Nikolay Borisov]

On 27.10.2017 10:31, Anand Jain wrote:
> btrfs_read_dev_super() returns -1 upon not finding a suitable SB,
> change that to return 1 instead, so that it can reserve the < 0
> values for the errno communications.
> 
> Signed-off-by: Anand Jain <anand.jain@oracle.com>

I believe this is buggy since it will be masking errors. We call
btrfs_read_dev_super in open_ctree_with_broken_chunk if the function
returns 1 we eventually exit as: return ERR_PTR(ret); And then perform
IS_ERR on the returned value. IS_ERR is essentially:
((x) >= (unsigned long)-MAX_ERRNO) where x could be 1 and MAX_ERRNO is 4095

So you just broke all functions which return the ret of
btrfs_read_dev_super as ERR_PTR.


NAK

> ---
> An independent patch, not related to any recent patch sent to ML.
> 
>  disk-io.c | 2 +-
>  utils.c   | 2 +-
>  volumes.c | 2 +-
>  3 files changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/disk-io.c b/disk-io.c
> index f5edc4796619..ba87bb4ce6da 100644
> --- a/disk-io.c
> +++ b/disk-io.c
> @@ -1491,7 +1491,7 @@ int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
>  		}
>  	}
>  
> -	return transid > 0 ? 0 : -1;
> +	return transid > 0 ? 0 : 1;
>  }
>  
>  static int write_dev_supers(struct btrfs_fs_info *fs_info,
> diff --git a/utils.c b/utils.c
> index 524f463d3140..7574cda8151c 100644
> --- a/utils.c
> +++ b/utils.c
> @@ -1678,7 +1678,7 @@ int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
>  		disk_super = (struct btrfs_super_block *)buf;
>  		ret = btrfs_read_dev_super(fd, disk_super,
>  					   BTRFS_SUPER_INFO_OFFSET, 0);
> -		if (ret < 0) {
> +		if (ret) {
>  			ret = -EIO;
>  			goto out;
>  		}
> diff --git a/volumes.c b/volumes.c
> index 2209e5a9100b..87008bfdd586 100644
> --- a/volumes.c
> +++ b/volumes.c
> @@ -268,7 +268,7 @@ int btrfs_scan_one_device(int fd, const char *path,
>  
>  	disk_super = (struct btrfs_super_block *)buf;
>  	ret = btrfs_read_dev_super(fd, disk_super, super_offset, sbflags);
> -	if (ret < 0)
> +	if (ret)
>  		return -EIO;
>  	devid = btrfs_stack_device_id(&disk_super->dev_item);
>  	if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_METADUMP)
> 

Re: [PATCH] btrfs-progs: reserve < 0 return value for errnos

[Qu Wenruo]

[-- Attachment #1.1: Type: text/plain, Size: 2808 bytes --]



On 2017年10月27日 15:54, Nikolay Borisov wrote:
> 
> 
> On 27.10.2017 10:31, Anand Jain wrote:
>> btrfs_read_dev_super() returns -1 upon not finding a suitable SB,
>> change that to return 1 instead, so that it can reserve the < 0
>> values for the errno communications.
>>
>> Signed-off-by: Anand Jain <anand.jain@oracle.com>
> 
> I believe this is buggy since it will be masking errors. We call
> btrfs_read_dev_super in open_ctree_with_broken_chunk if the function
> returns 1 we eventually exit as: return ERR_PTR(ret); And then perform
> IS_ERR on the returned value. IS_ERR is essentially:
> ((x) >= (unsigned long)-MAX_ERRNO) where x could be 1 and MAX_ERRNO is 4095
> 
> So you just broke all functions which return the ret of
> btrfs_read_dev_super as ERR_PTR.
> 
> 
> NAK

This doesn't looks good to me either.

No suitable superblock is already an error.
Why not just returning -ENOENT and let caller who really needs to
distinguish this to catch that -ENOENT?

Thanks,
Qu

> 
>> ---
>> An independent patch, not related to any recent patch sent to ML.
>>
>>  disk-io.c | 2 +-
>>  utils.c   | 2 +-
>>  volumes.c | 2 +-
>>  3 files changed, 3 insertions(+), 3 deletions(-)
>>
>> diff --git a/disk-io.c b/disk-io.c
>> index f5edc4796619..ba87bb4ce6da 100644
>> --- a/disk-io.c
>> +++ b/disk-io.c
>> @@ -1491,7 +1491,7 @@ int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
>>  		}
>>  	}
>>  
>> -	return transid > 0 ? 0 : -1;
>> +	return transid > 0 ? 0 : 1;
>>  }
>>  
>>  static int write_dev_supers(struct btrfs_fs_info *fs_info,
>> diff --git a/utils.c b/utils.c
>> index 524f463d3140..7574cda8151c 100644
>> --- a/utils.c
>> +++ b/utils.c
>> @@ -1678,7 +1678,7 @@ int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
>>  		disk_super = (struct btrfs_super_block *)buf;
>>  		ret = btrfs_read_dev_super(fd, disk_super,
>>  					   BTRFS_SUPER_INFO_OFFSET, 0);
>> -		if (ret < 0) {
>> +		if (ret) {
>>  			ret = -EIO;
>>  			goto out;
>>  		}
>> diff --git a/volumes.c b/volumes.c
>> index 2209e5a9100b..87008bfdd586 100644
>> --- a/volumes.c
>> +++ b/volumes.c
>> @@ -268,7 +268,7 @@ int btrfs_scan_one_device(int fd, const char *path,
>>  
>>  	disk_super = (struct btrfs_super_block *)buf;
>>  	ret = btrfs_read_dev_super(fd, disk_super, super_offset, sbflags);
>> -	if (ret < 0)
>> +	if (ret)
>>  		return -EIO;
>>  	devid = btrfs_stack_device_id(&disk_super->dev_item);
>>  	if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_METADUMP)
>>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 520 bytes --]

Re: [PATCH] btrfs-progs: reserve < 0 return value for errnos

[Anand Jain]

On 10/27/2017 04:04 PM, Qu Wenruo wrote:
> 
> 
> On 2017年10月27日 15:54, Nikolay Borisov wrote:
>>
>>
>> On 27.10.2017 10:31, Anand Jain wrote:
>>> btrfs_read_dev_super() returns -1 upon not finding a suitable SB,
>>> change that to return 1 instead, so that it can reserve the < 0
>>> values for the errno communications.
>>>
>>> Signed-off-by: Anand Jain <anand.jain@oracle.com>
>>
>> I believe this is buggy since it will be masking errors. We call
>> btrfs_read_dev_super in open_ctree_with_broken_chunk if the function
>> returns 1 we eventually exit as: return ERR_PTR(ret); And then perform
>> IS_ERR on the returned value. IS_ERR is essentially:
>> ((x) >= (unsigned long)-MAX_ERRNO) where x could be 1 and MAX_ERRNO is 4095
>>
>> So you just broke all functions which return the ret of
>> btrfs_read_dev_super as ERR_PTR.
>>
>>
>> NAK
> 
> This doesn't looks good to me either.
> 
> No suitable superblock is already an error.
> Why not just returning -ENOENT and let caller who really needs to
> distinguish this to catch that -ENOENT?

  Thanks for pointing out. Will fix it. IMO, -ENOENT is good for now.

  (Though the actual cleanup would be to..
     There are two parts in btrfs_read_dev_super() one to read
     only the BTRFS_SUPER_INFO_OFFSET SB. And the other to pick
     the latest SB by generation #. So splitting them two separate
     functions will let us know where do we use what in a much
     clearer way. And for 2nd pread64 we should probably use errno.)

Thanks, Anand


> Thanks,
> Qu
> 
>>
>>> ---
>>> An independent patch, not related to any recent patch sent to ML.
>>>
>>>   disk-io.c | 2 +-
>>>   utils.c   | 2 +-
>>>   volumes.c | 2 +-
>>>   3 files changed, 3 insertions(+), 3 deletions(-)
>>>
>>> diff --git a/disk-io.c b/disk-io.c
>>> index f5edc4796619..ba87bb4ce6da 100644
>>> --- a/disk-io.c
>>> +++ b/disk-io.c
>>> @@ -1491,7 +1491,7 @@ int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
>>>   		}
>>>   	}
>>>   
>>> -	return transid > 0 ? 0 : -1;
>>> +	return transid > 0 ? 0 : 1;
>>>   }
>>>   
>>>   static int write_dev_supers(struct btrfs_fs_info *fs_info,
>>> diff --git a/utils.c b/utils.c
>>> index 524f463d3140..7574cda8151c 100644
>>> --- a/utils.c
>>> +++ b/utils.c
>>> @@ -1678,7 +1678,7 @@ int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
>>>   		disk_super = (struct btrfs_super_block *)buf;
>>>   		ret = btrfs_read_dev_super(fd, disk_super,
>>>   					   BTRFS_SUPER_INFO_OFFSET, 0);
>>> -		if (ret < 0) {
>>> +		if (ret) {
>>>   			ret = -EIO;
>>>   			goto out;
>>>   		}
>>> diff --git a/volumes.c b/volumes.c
>>> index 2209e5a9100b..87008bfdd586 100644
>>> --- a/volumes.c
>>> +++ b/volumes.c
>>> @@ -268,7 +268,7 @@ int btrfs_scan_one_device(int fd, const char *path,
>>>   
>>>   	disk_super = (struct btrfs_super_block *)buf;
>>>   	ret = btrfs_read_dev_super(fd, disk_super, super_offset, sbflags);
>>> -	if (ret < 0)
>>> +	if (ret)
>>>   		return -EIO;
>>>   	devid = btrfs_stack_device_id(&disk_super->dev_item);
>>>   	if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_METADUMP)
>>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>
> 

[PATCH v2] btrfs-progs: reserve < 0 return value for errnos

[Anand Jain]

btrfs_read_dev_super() returns -1 / -EPERM upon not finding a suitable SB,
change that to return -ENOENT instead, which is much closer to the actual
possible errors.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
---
v2: use -ENOENT instead of 1.

 disk-io.c | 2 +-
 utils.c   | 2 +-
 volumes.c | 2 +-
 3 files changed, 3 insertions(+), 3 deletions(-)

diff --git a/disk-io.c b/disk-io.c
index f5edc4796619..dcaaea898a3f 100644
--- a/disk-io.c
+++ b/disk-io.c
@@ -1491,7 +1491,7 @@ int btrfs_read_dev_super(int fd, struct btrfs_super_block *sb, u64 sb_bytenr,
 		}
 	}
 
-	return transid > 0 ? 0 : -1;
+	return transid > 0 ? 0 : -ENOENT;
 }
 
 static int write_dev_supers(struct btrfs_fs_info *fs_info,
diff --git a/utils.c b/utils.c
index 524f463d3140..7574cda8151c 100644
--- a/utils.c
+++ b/utils.c
@@ -1678,7 +1678,7 @@ int get_fs_info(const char *path, struct btrfs_ioctl_fs_info_args *fi_args,
 		disk_super = (struct btrfs_super_block *)buf;
 		ret = btrfs_read_dev_super(fd, disk_super,
 					   BTRFS_SUPER_INFO_OFFSET, 0);
-		if (ret < 0) {
+		if (ret) {
 			ret = -EIO;
 			goto out;
 		}
diff --git a/volumes.c b/volumes.c
index 2209e5a9100b..87008bfdd586 100644
--- a/volumes.c
+++ b/volumes.c
@@ -268,7 +268,7 @@ int btrfs_scan_one_device(int fd, const char *path,
 
 	disk_super = (struct btrfs_super_block *)buf;
 	ret = btrfs_read_dev_super(fd, disk_super, super_offset, sbflags);
-	if (ret < 0)
+	if (ret)
 		return -EIO;
 	devid = btrfs_stack_device_id(&disk_super->dev_item);
 	if (btrfs_super_flags(disk_super) & BTRFS_SUPER_FLAG_METADUMP)
-- 
2.13.1