[PATCH] vfs: added better file aio_read aio_write operations presence check

[PATCH] vfs: added better file aio_read aio_write operations presence check

[Marco Stornelli]

From: Marco Stornelli <marco.stornelli@gmail.com>

If a filesystem in the file operations specifies for read and write operations only do_sync_read and do_sync_write without 
init aio_read and aio_write, there will be a kernel oops, because the vfs code check the presence of (to read for example)
read OR aio_read method, then it calls read if it's pointer is not null. It's not sufficient because if the read function is 
actually a do_sync_read, it calls aio_read but without checking the presence. I think a BUG_ON check can be more useful.
Signed-off-by: Marco Stornelli <marco.stornelli@gmail.com>
---

--- linux-2.6.26.5/fs/read_write.c.orig	2008-08-20 20:11:37.000000000 +0200
+++ linux-2.6.26.5/fs/read_write.c	2008-09-16 11:01:13.000000000 +0200
@@ -240,6 +240,7 @@ ssize_t do_sync_read(struct file *filp,
 	kiocb.ki_pos = *ppos;
 	kiocb.ki_left = len;
 
+	BUG_ON(!filp->f_op->aio_read);
 	for (;;) {
 		ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
 		if (ret != -EIOCBRETRY)
@@ -295,6 +296,7 @@ ssize_t do_sync_write(struct file *filp,
 	kiocb.ki_pos = *ppos;
 	kiocb.ki_left = len;
 
+	BUG_ON(!filp->f_op->aio_write);
 	for (;;) {
 		ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
 		if (ret != -EIOCBRETRY)

Re: [PATCH] vfs: added better file aio_read aio_write operations presence check

[Manish Katiyar]

On Tue, Sep 16, 2008 at 2:59 PM, Marco Stornelli
<marco.stornelli@coritel.it> wrote:
> From: Marco Stornelli <marco.stornelli@gmail.com>
>
> If a filesystem in the file operations specifies for read and write operations only do_sync_read and do_sync_write without
> init aio_read and aio_write, there will be a kernel oops, because the vfs code check the presence of (to read for example)
> read OR aio_read method, then it calls read if it's pointer is not null. It's not sufficient because if the read function is
> actually a do_sync_read, it calls aio_read but without checking the presence. I think a BUG_ON check can be more useful.

Instead of doing a BUG_ON() why can't we simply fall back to the
generic_aio functions since most of the fs tend to do so as below.


Signed-off-by: Manish Katiyar <mkatiyar@gmail.com>

---
 fs/read_write.c |   10 ++++++++--
 1 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/fs/read_write.c b/fs/read_write.c
index 9ba495d..5439bc4 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -225,7 +225,11 @@ ssize_t do_sync_read(struct file *filp, char
__user *buf, size_t len, loff_t *pp
 	kiocb.ki_left = len;

 	for (;;) {
-		ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
+		if (filp->f_op->aio_read)
+			ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
+		else
+			ret = generic_file_aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
 		if (ret != -EIOCBRETRY)
 			break;
 		wait_on_retry_sync_kiocb(&kiocb);
@@ -280,7 +284,10 @@ ssize_t do_sync_write(struct file *filp, const
char __user *buf, size_t len, lof
 	kiocb.ki_left = len;

 	for (;;) {
-		ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
+		if (filp->f_op->aio_write)
+			ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
+		else
+			ret = generic_file_aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
 		if (ret != -EIOCBRETRY)
 			break;
 		wait_on_retry_sync_kiocb(&kiocb);
-- 
1.5.4.3


Thanks -
Manish

> Signed-off-by: Marco Stornelli <marco.stornelli@gmail.com>
> ---
>
> --- linux-2.6.26.5/fs/read_write.c.orig 2008-08-20 20:11:37.000000000 +0200
> +++ linux-2.6.26.5/fs/read_write.c      2008-09-16 11:01:13.000000000 +0200
> @@ -240,6 +240,7 @@ ssize_t do_sync_read(struct file *filp,
>        kiocb.ki_pos = *ppos;
>        kiocb.ki_left = len;
>
> +       BUG_ON(!filp->f_op->aio_read);
>        for (;;) {
>                ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
>                if (ret != -EIOCBRETRY)
> @@ -295,6 +296,7 @@ ssize_t do_sync_write(struct file *filp,
>        kiocb.ki_pos = *ppos;
>        kiocb.ki_left = len;
>
> +       BUG_ON(!filp->f_op->aio_write);
>        for (;;) {
>                ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
>                if (ret != -EIOCBRETRY)
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

Re: [PATCH] vfs: added better file aio_read aio_write operations presence check

[Marco Stornelli]

BUG_ON it was a way to say: "hey you've used the do_sync_read/write as
read/write operation but you don't specified an aio_read/write", but
your solutions it's good too.

Manish Katiyar ha scritto:
> On Tue, Sep 16, 2008 at 2:59 PM, Marco Stornelli
> <marco.stornelli@coritel.it> wrote:
>> From: Marco Stornelli <marco.stornelli@gmail.com>
>>
>> If a filesystem in the file operations specifies for read and write operations only do_sync_read and do_sync_write without
>> init aio_read and aio_write, there will be a kernel oops, because the vfs code check the presence of (to read for example)
>> read OR aio_read method, then it calls read if it's pointer is not null. It's not sufficient because if the read function is
>> actually a do_sync_read, it calls aio_read but without checking the presence. I think a BUG_ON check can be more useful.
> 
> Instead of doing a BUG_ON() why can't we simply fall back to the
> generic_aio functions since most of the fs tend to do so as below.
> 
> 
> Signed-off-by: Manish Katiyar <mkatiyar@gmail.com>
> 
> ---
>  fs/read_write.c |   10 ++++++++--
>  1 files changed, 8 insertions(+), 2 deletions(-)
> 
> diff --git a/fs/read_write.c b/fs/read_write.c
> index 9ba495d..5439bc4 100644
> --- a/fs/read_write.c
> +++ b/fs/read_write.c
> @@ -225,7 +225,11 @@ ssize_t do_sync_read(struct file *filp, char
> __user *buf, size_t len, loff_t *pp
>  	kiocb.ki_left = len;
> 
>  	for (;;) {
> -		ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
> +		if (filp->f_op->aio_read)
> +			ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
> +		else
> +			ret = generic_file_aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
>  		if (ret != -EIOCBRETRY)
>  			break;
>  		wait_on_retry_sync_kiocb(&kiocb);
> @@ -280,7 +284,10 @@ ssize_t do_sync_write(struct file *filp, const
> char __user *buf, size_t len, lof
>  	kiocb.ki_left = len;
> 
>  	for (;;) {
> -		ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
> +		if (filp->f_op->aio_write)
> +			ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
> +		else
> +			ret = generic_file_aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
>  		if (ret != -EIOCBRETRY)
>  			break;
>  		wait_on_retry_sync_kiocb(&kiocb);

Re: [PATCH] vfs: added better file aio_read aio_write operations presence check

[Manish Katiyar]

On Tue, Sep 16, 2008 at 4:43 PM, Marco Stornelli
<marco.stornelli@coritel.it> wrote:
> BUG_ON it was a way to say: "hey you've used the do_sync_read/write as
> read/write operation but you don't specified an aio_read/write", but
> your solutions it's good too.

Looks like I made some copy paste error while sending the patch. Below
is the updated one.

Signed-off-by: Manish Katiyar <mkatiyar@gmail.com>

---
 fs/read_write.c |   10 ++++++++--
 1 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/fs/read_write.c b/fs/read_write.c
index 9ba495d..b89b707 100644
--- a/fs/read_write.c
+++ b/fs/read_write.c
@@ -225,7 +225,10 @@ ssize_t do_sync_read(struct file *filp, char
__user *buf, size_t len, loff_t *pp
 	kiocb.ki_left = len;

 	for (;;) {
-		ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
+		if (filp->f_op->aio_read)
+			ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
+		else
+			ret = generic_file_aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
 		if (ret != -EIOCBRETRY)
 			break;
 		wait_on_retry_sync_kiocb(&kiocb);
@@ -280,7 +283,10 @@ ssize_t do_sync_write(struct file *filp, const
char __user *buf, size_t len, lof
 	kiocb.ki_left = len;

 	for (;;) {
-		ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
+		if (filp->f_op->aio_write)
+			ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
+		else
+			ret = generic_file_aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
 		if (ret != -EIOCBRETRY)
 			break;
 		wait_on_retry_sync_kiocb(&kiocb);
-- 
1.5.4.3

Thanks -
Manish


>
> Manish Katiyar ha scritto:
>> On Tue, Sep 16, 2008 at 2:59 PM, Marco Stornelli
>> <marco.stornelli@coritel.it> wrote:
>>> From: Marco Stornelli <marco.stornelli@gmail.com>
>>>
>>> If a filesystem in the file operations specifies for read and write operations only do_sync_read and do_sync_write without
>>> init aio_read and aio_write, there will be a kernel oops, because the vfs code check the presence of (to read for example)
>>> read OR aio_read method, then it calls read if it's pointer is not null. It's not sufficient because if the read function is
>>> actually a do_sync_read, it calls aio_read but without checking the presence. I think a BUG_ON check can be more useful.
>>
>> Instead of doing a BUG_ON() why can't we simply fall back to the
>> generic_aio functions since most of the fs tend to do so as below.
>>
>>
>> Signed-off-by: Manish Katiyar <mkatiyar@gmail.com>
>>
>> ---
>>  fs/read_write.c |   10 ++++++++--
>>  1 files changed, 8 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/read_write.c b/fs/read_write.c
>> index 9ba495d..5439bc4 100644
>> --- a/fs/read_write.c
>> +++ b/fs/read_write.c
>> @@ -225,7 +225,11 @@ ssize_t do_sync_read(struct file *filp, char
>> __user *buf, size_t len, loff_t *pp
>>       kiocb.ki_left = len;
>>
>>       for (;;) {
>> -             ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
>> +             if (filp->f_op->aio_read)
>> +                     ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
>> +             else
>> +                     ret = generic_file_aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
>>               if (ret != -EIOCBRETRY)
>>                       break;
>>               wait_on_retry_sync_kiocb(&kiocb);
>> @@ -280,7 +284,10 @@ ssize_t do_sync_write(struct file *filp, const
>> char __user *buf, size_t len, lof
>>       kiocb.ki_left = len;
>>
>>       for (;;) {
>> -             ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
>> +             if (filp->f_op->aio_write)
>> +                     ret = filp->f_op->aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
>> +             else
>> +                     ret = generic_file_aio_write(&kiocb, &iov, 1, kiocb.ki_pos);
>>               if (ret != -EIOCBRETRY)
>>                       break;
>>               wait_on_retry_sync_kiocb(&kiocb);
>
>

Re: [PATCH] vfs: added better file aio_read aio_write operations presence check

[Bodo Eggert]

Manish Katiyar <mkatiyar@gmail.com> wrote:
> On Tue, Sep 16, 2008 at 2:59 PM, Marco Stornelli
>> From: Marco Stornelli <marco.stornelli@gmail.com>

>> If a filesystem in the file operations specifies for read and write
>> operations only do_sync_read and do_sync_write without init aio_read and
>> aio_write, there will be a kernel oops, because the vfs code check the
>> presence of (to read for example) read OR aio_read method, then it calls read
>> if it's pointer is not null. It's not sufficient because if the read function
>> is actually a do_sync_read, it calls aio_read but without checking the
>> presence. I think a BUG_ON check can be more useful.
> 
> Instead of doing a BUG_ON() why can't we simply fall back to the
> generic_aio functions since most of the fs tend to do so as below.

> --- a/fs/read_write.c

> -             ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
> +             if (filp->f_op->aio_read)
> +                     ret = filp->f_op->aio_read(&kiocb, &iov, 1, kiocb.ki_pos);
> +             else
> +                     ret = generic_file_aio_read(&kiocb, &iov, 1, kiocb.ki_pos);

Why can't the file system registration code set filp->f_op->aio_read to
generic_file_aio_read?

Re: [PATCH] vfs: added better file aio_read aio_write operations presence check

[Matthew Wilcox]

On Tue, Sep 16, 2008 at 05:36:09PM +0200, Bodo Eggert wrote:
> Why can't the file system registration code set filp->f_op->aio_read to
> generic_file_aio_read?

        const struct file_operations    *f_op;

Having said that, BUILD_BUG_ON(!f_op->aio_read) is fine by me ... make
the filesystem writer put it in without slowing down anyone at runtime.

-- 
Matthew Wilcox				Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."

Re: [PATCH] vfs: added better file aio_read aio_write operations presence check

[Christoph Hellwig]

On Tue, Sep 16, 2008 at 11:29:41AM +0200, Marco Stornelli wrote:
> From: Marco Stornelli <marco.stornelli@gmail.com>
> 
> If a filesystem in the file operations specifies for read and write operations only do_sync_read and do_sync_write without 
> init aio_read and aio_write, there will be a kernel oops, because the vfs code check the presence of (to read for example)
> read OR aio_read method, then it calls read if it's pointer is not null. It's not sufficient because if the read function is 
> actually a do_sync_read, it calls aio_read but without checking the presence. I think a BUG_ON check can be more useful.

A NULL pointer derference is just as clear as the bug..