[PATCH] cifs: small underflow in cnvrtDosUnixTm()

[PATCH] cifs: small underflow in cnvrtDosUnixTm()

[Dan Carpenter]

January is month 1.  There is no zero-th month.  We don't care very much
if the days are invalid but for months, we use it to read from an array
so this bug means we read one space before the start of the
total_days_of_prev_months[] array.

Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
Signed-off-by: Dan Carpenter <dan.carpenter-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>

diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
index abae6dd2c6b9..f1f64a15215a 100644
--- a/fs/cifs/netmisc.c
+++ b/fs/cifs/netmisc.c
@@ -980,8 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
 		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
 	days = sd->Day;
 	month = sd->Month;
-	if ((days > 31) || (month > 12)) {
+	if (days > 31 || month < 1 || month > 12) {
 		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
+		if (month < 1)
+			month = 1;
 		if (month > 12)
 			month = 12;
 	}

[PATCH] cifs: small underflow in cnvrtDosUnixTm()

[Dan Carpenter]

January is month 1.  There is no zero-th month.  We don't care very much
if the days are invalid but for months, we use it to read from an array
so this bug means we read one space before the start of the
total_days_of_prev_months[] array.

Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>

diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
index abae6dd2c6b9..f1f64a15215a 100644
--- a/fs/cifs/netmisc.c
+++ b/fs/cifs/netmisc.c
@@ -980,8 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
 		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
 	days = sd->Day;
 	month = sd->Month;
-	if ((days > 31) || (month > 12)) {
+	if (days > 31 || month < 1 || month > 12) {
 		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
+		if (month < 1)
+			month = 1;
 		if (month > 12)
 			month = 12;
 	}

[PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[Dan Carpenter]

January is month 1.  There is no zero-th month.  If someone passes a
zero month then it means we read from one space before the start of the
total_days_of_prev_months[] array.

We may as well also be strict about days as well.

Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
Signed-off-by: Dan Carpenter <dan.carpenter-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
---
v2:  Be strict about days as well.  My first patch was less intrusive
because it only prevented the out of bounds access.  I have no idea how
common it is to pass in an illegal day but, hopefully, not very common.

diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
index abae6dd2c6b9..4b2726ee4fad 100644
--- a/fs/cifs/netmisc.c
+++ b/fs/cifs/netmisc.c
@@ -980,10 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
 		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
 	days = sd->Day;
 	month = sd->Month;
-	if ((days > 31) || (month > 12)) {
+	if (days < 1 || days > 31 || month < 1 || month > 12) {
 		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
-		if (month > 12)
-			month = 12;
+		days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
+		month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
 	}
 	month -= 1;
 	days += total_days_of_prev_months[month];

[PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[Dan Carpenter]

January is month 1.  There is no zero-th month.  If someone passes a
zero month then it means we read from one space before the start of the
total_days_of_prev_months[] array.

We may as well also be strict about days as well.

Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
---
v2:  Be strict about days as well.  My first patch was less intrusive
because it only prevented the out of bounds access.  I have no idea how
common it is to pass in an illegal day but, hopefully, not very common.

diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
index abae6dd2c6b9..4b2726ee4fad 100644
--- a/fs/cifs/netmisc.c
+++ b/fs/cifs/netmisc.c
@@ -980,10 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
 		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
 	days = sd->Day;
 	month = sd->Month;
-	if ((days > 31) || (month > 12)) {
+	if (days < 1 || days > 31 || month < 1 || month > 12) {
 		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
-		if (month > 12)
-			month = 12;
+		days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
+		month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
 	}
 	month -= 1;
 	days += total_days_of_prev_months[month];

Re: [PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[walter harms]

Am 28.04.2017 14:51, schrieb Dan Carpenter:
> January is month 1.  There is no zero-th month.  If someone passes a
> zero month then it means we read from one space before the start of the
> total_days_of_prev_months[] array.
> 
> We may as well also be strict about days as well.
> 
> Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> ---
> v2:  Be strict about days as well.  My first patch was less intrusive
> because it only prevented the out of bounds access.  I have no idea how
> common it is to pass in an illegal day but, hopefully, not very common.
> 
> diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
> index abae6dd2c6b9..4b2726ee4fad 100644
> --- a/fs/cifs/netmisc.c
> +++ b/fs/cifs/netmisc.c
> @@ -980,10 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
>  		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
>  	days = sd->Day;
>  	month = sd->Month;
> -	if ((days > 31) || (month > 12)) {
> +	if (days < 1 || days > 31 || month < 1 || month > 12) {
>  		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
> -		if (month > 12)
> -			month = 12;
> +		days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
> +		month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
>  	}
>  	month -= 1;
>  	days += total_days_of_prev_months[month];

The mixing in now a bit unfortunate ... why not simply

	if (days < 1 || days > 31 || month < 1 || month > 12)
		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);

	month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
	days = (days < 1) ? 1 : ((days <= 31) ? days : 31);

	
hope that helps,
re,
 wh


> --
> To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: [PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[walter harms]

Am 28.04.2017 14:51, schrieb Dan Carpenter:
> January is month 1.  There is no zero-th month.  If someone passes a
> zero month then it means we read from one space before the start of the
> total_days_of_prev_months[] array.
> 
> We may as well also be strict about days as well.
> 
> Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> ---
> v2:  Be strict about days as well.  My first patch was less intrusive
> because it only prevented the out of bounds access.  I have no idea how
> common it is to pass in an illegal day but, hopefully, not very common.
> 
> diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
> index abae6dd2c6b9..4b2726ee4fad 100644
> --- a/fs/cifs/netmisc.c
> +++ b/fs/cifs/netmisc.c
> @@ -980,10 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
>  		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
>  	days = sd->Day;
>  	month = sd->Month;
> -	if ((days > 31) || (month > 12)) {
> +	if (days < 1 || days > 31 || month < 1 || month > 12) {
>  		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
> -		if (month > 12)
> -			month = 12;
> +		days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
> +		month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
>  	}
>  	month -= 1;
>  	days += total_days_of_prev_months[month];

The mixing in now a bit unfortunate ... why not simply

	if (days < 1 || days > 31 || month < 1 || month > 12)
		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);

	month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
	days = (days < 1) ? 1 : ((days <= 31) ? days : 31);

	
hope that helps,
re,
 wh


> --
> To unsubscribe from this list: send the line "unsubscribe kernel-janitors" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: [PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[Dan Carpenter]

On Fri, Apr 28, 2017 at 04:40:04PM +0200, walter harms wrote:
> 
> 
> Am 28.04.2017 14:51, schrieb Dan Carpenter:
> > January is month 1.  There is no zero-th month.  If someone passes a
> > zero month then it means we read from one space before the start of the
> > total_days_of_prev_months[] array.
> > 
> > We may as well also be strict about days as well.
> > 
> > Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
> > Signed-off-by: Dan Carpenter <dan.carpenter-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
> > ---
> > v2:  Be strict about days as well.  My first patch was less intrusive
> > because it only prevented the out of bounds access.  I have no idea how
> > common it is to pass in an illegal day but, hopefully, not very common.
> > 
> > diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
> > index abae6dd2c6b9..4b2726ee4fad 100644
> > --- a/fs/cifs/netmisc.c
> > +++ b/fs/cifs/netmisc.c
> > @@ -980,10 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
> >  		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
> >  	days = sd->Day;
> >  	month = sd->Month;
> > -	if ((days > 31) || (month > 12)) {
> > +	if (days < 1 || days > 31 || month < 1 || month > 12) {
> >  		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
> > -		if (month > 12)
> > -			month = 12;
> > +		days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
> > +		month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
> >  	}
> >  	month -= 1;
> >  	days += total_days_of_prev_months[month];
> 
> The mixing in now a bit unfortunate ... why not simply
> 
> 	if (days < 1 || days > 31 || month < 1 || month > 12)
> 		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
> 
> 	month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
> 	days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
> 

I prefer my version because I feel like it more closely expresses what
I want to say.

regards,
dan carpenter

Re: [PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[Dan Carpenter]

On Fri, Apr 28, 2017 at 04:40:04PM +0200, walter harms wrote:
> 
> 
> Am 28.04.2017 14:51, schrieb Dan Carpenter:
> > January is month 1.  There is no zero-th month.  If someone passes a
> > zero month then it means we read from one space before the start of the
> > total_days_of_prev_months[] array.
> > 
> > We may as well also be strict about days as well.
> > 
> > Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
> > Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
> > ---
> > v2:  Be strict about days as well.  My first patch was less intrusive
> > because it only prevented the out of bounds access.  I have no idea how
> > common it is to pass in an illegal day but, hopefully, not very common.
> > 
> > diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
> > index abae6dd2c6b9..4b2726ee4fad 100644
> > --- a/fs/cifs/netmisc.c
> > +++ b/fs/cifs/netmisc.c
> > @@ -980,10 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
> >  		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
> >  	days = sd->Day;
> >  	month = sd->Month;
> > -	if ((days > 31) || (month > 12)) {
> > +	if (days < 1 || days > 31 || month < 1 || month > 12) {
> >  		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
> > -		if (month > 12)
> > -			month = 12;
> > +		days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
> > +		month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
> >  	}
> >  	month -= 1;
> >  	days += total_days_of_prev_months[month];
> 
> The mixing in now a bit unfortunate ... why not simply
> 
> 	if (days < 1 || days > 31 || month < 1 || month > 12)
> 		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
> 
> 	month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
> 	days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
> 

I prefer my version because I feel like it more closely expresses what
I want to say.

regards,
dan carpenter

Re: [PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[walter harms]

Am 28.04.2017 16:41, schrieb Dan Carpenter:
> On Fri, Apr 28, 2017 at 04:40:04PM +0200, walter harms wrote:
>>
>>
>> Am 28.04.2017 14:51, schrieb Dan Carpenter:
>>> January is month 1.  There is no zero-th month.  If someone passes a
>>> zero month then it means we read from one space before the start of the
>>> total_days_of_prev_months[] array.
>>>
>>> We may as well also be strict about days as well.
>>>
>>> Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
>>> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
>>> ---
>>> v2:  Be strict about days as well.  My first patch was less intrusive
>>> because it only prevented the out of bounds access.  I have no idea how
>>> common it is to pass in an illegal day but, hopefully, not very common.
>>>
>>> diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
>>> index abae6dd2c6b9..4b2726ee4fad 100644
>>> --- a/fs/cifs/netmisc.c
>>> +++ b/fs/cifs/netmisc.c
>>> @@ -980,10 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
>>>  		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
>>>  	days = sd->Day;
>>>  	month = sd->Month;
>>> -	if ((days > 31) || (month > 12)) {
>>> +	if (days < 1 || days > 31 || month < 1 || month > 12) {
>>>  		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
>>> -		if (month > 12)
>>> -			month = 12;
>>> +		days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
>>> +		month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
>>>  	}
>>>  	month -= 1;
>>>  	days += total_days_of_prev_months[month];
>>
>> The mixing in now a bit unfortunate ... why not simply
>>
>> 	if (days < 1 || days > 31 || month < 1 || month > 12)
>> 		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
>>
>> 	month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
>> 	days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
>>
> 
> I prefer my version because I feel like it more closely expresses what
> I want to say.
> 

(Actually it is your code just without the braces.)
what is about differentiating between day and month ?

if (days < 1 || days > 31) {
		cifs_dbg(VFS, "illegal day: %d\n", days);
		days = (days < 1) ? 1 : 31 ;
		}


if ( month < 1 || month > 12) {
		cifs_dbg(VFS, "illegal month %d\n", month);
		month = (month < 1) ? 1 : 12 ;
		}

this way you have a obvious error handling for day and month.

I still like my first version even when day/month are recalculated every
time the flow is very linear only the error message is a deviation.
(Given that ?: makes it not easy to digest)

re,
 wh

Re: [PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[walter harms]

Am 28.04.2017 16:41, schrieb Dan Carpenter:
> On Fri, Apr 28, 2017 at 04:40:04PM +0200, walter harms wrote:
>>
>>
>> Am 28.04.2017 14:51, schrieb Dan Carpenter:
>>> January is month 1.  There is no zero-th month.  If someone passes a
>>> zero month then it means we read from one space before the start of the
>>> total_days_of_prev_months[] array.
>>>
>>> We may as well also be strict about days as well.
>>>
>>> Fixes: 1bd5bbcb6531 ("[CIFS] Legacy time handling for Win9x and OS/2 part 1")
>>> Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
>>> ---
>>> v2:  Be strict about days as well.  My first patch was less intrusive
>>> because it only prevented the out of bounds access.  I have no idea how
>>> common it is to pass in an illegal day but, hopefully, not very common.
>>>
>>> diff --git a/fs/cifs/netmisc.c b/fs/cifs/netmisc.c
>>> index abae6dd2c6b9..4b2726ee4fad 100644
>>> --- a/fs/cifs/netmisc.c
>>> +++ b/fs/cifs/netmisc.c
>>> @@ -980,10 +980,10 @@ struct timespec cnvrtDosUnixTm(__le16 le_date, __le16 le_time, int offset)
>>>  		cifs_dbg(VFS, "illegal hours %d\n", st->Hours);
>>>  	days = sd->Day;
>>>  	month = sd->Month;
>>> -	if ((days > 31) || (month > 12)) {
>>> +	if (days < 1 || days > 31 || month < 1 || month > 12) {
>>>  		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
>>> -		if (month > 12)
>>> -			month = 12;
>>> +		days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
>>> +		month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
>>>  	}
>>>  	month -= 1;
>>>  	days += total_days_of_prev_months[month];
>>
>> The mixing in now a bit unfortunate ... why not simply
>>
>> 	if (days < 1 || days > 31 || month < 1 || month > 12)
>> 		cifs_dbg(VFS, "illegal date, month %d day: %d\n", month, days);
>>
>> 	month = (month < 1) ? 1 : ((month <= 12) ? month : 12);
>> 	days = (days < 1) ? 1 : ((days <= 31) ? days : 31);
>>
> 
> I prefer my version because I feel like it more closely expresses what
> I want to say.
> 

(Actually it is your code just without the braces.)
what is about differentiating between day and month ?

if (days < 1 || days > 31) {
		cifs_dbg(VFS, "illegal day: %d\n", days);
		days = (days < 1) ? 1 : 31 ;
		}


if ( month < 1 || month > 12) {
		cifs_dbg(VFS, "illegal month %d\n", month);
		month = (month < 1) ? 1 : 12 ;
		}

this way you have a obvious error handling for day and month.

I still like my first version even when day/month are recalculated every
time the flow is very linear only the error message is a deviation.
(Given that ?: makes it not easy to digest)

re,
 wh

Re: [PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[Dan Carpenter]

I will resend on Monday.

regards,
dan carpenter

Re: [PATCH v2] cifs: small underflow in cnvrtDosUnixTm()

[Dan Carpenter]

I will resend on Monday.

regards,
dan carpenter