Re: [PATCH] SCTP: Bring SCTP_DELAYED_ACK socket option into ietf

[Vlad Yasevich <vladislav.yasevich@hp.com>]

Hi Wei

Some comments.

Wei Yongjun wrote:
> Brings delayed_ack socket option set/get into line with the latest ietf 
> socket extensions API draft, while maintaining backwards compatibility.
> 
> Signed-off-by: Wei Yongjun <yjwei@cn.fujitsu.com>
> 
> --- a/include/net/sctp/user.h    2008-04-18 13:45:11.000000000 -0400
> +++ b/include/net/sctp/user.h    2008-04-19 08:36:05.000000000 -0400
> @@ -93,8 +93,8 @@ enum sctp_optname {
> #define SCTP_STATUS SCTP_STATUS
>     SCTP_GET_PEER_ADDR_INFO,
> #define SCTP_GET_PEER_ADDR_INFO SCTP_GET_PEER_ADDR_INFO
> -    SCTP_DELAYED_ACK_TIME,
> -#define SCTP_DELAYED_ACK_TIME SCTP_DELAYED_ACK_TIME
> +    SCTP_DELAYED_ACK,
> +#define SCTP_DELAYED_ACK SCTP_DELAYED_ACK
>     SCTP_CONTEXT,    /* Receive Context */
> #define SCTP_CONTEXT SCTP_CONTEXT
>     SCTP_FRAGMENT_INTERLEAVE,
> @@ -618,16 +618,29 @@ struct sctp_authkeyid {
> };
> 

If you are going to do this, you need to leave the old define symbol around.
That way, applications using the symbol will still compile.

> 
> -/* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
> +/*
> + * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
>  *
> - *   This options will get or set the delayed ack timer.  The time is set
> - *   in milliseconds.  If the assoc_id is 0, then this sets or gets the
> - *   endpoints default delayed ack timer value.  If the assoc_id field is
> - *   non-zero, then the set or get effects the specified association.
> - */
> + * This option will effect the way delayed acks are performed.  This
> + * option allows you to get or set the delayed ack time, in
> + * milliseconds.  It also allows changing the delayed ack frequency.
> + * Changing the frequency to 1 disables the delayed sack algorithm.  If
> + * the assoc_id is 0, then this sets or gets the endpoints default
> + * values.  If the assoc_id field is non-zero, then the set or get
> + * effects the specified association for the one to many model (the
> + * assoc_id field is ignored by the one to one model).  Note that if
> + * sack_delay or sack_freq are 0 when setting this option, then the
> + * current values will remain unchanged.
> + */
> +struct sctp_sack_info {
> +    sctp_assoc_t    sack_assoc_id;
> +    uint32_t    sack_delay;
> +    uint32_t    sack_freq;
> +};
> +
> struct sctp_assoc_value {
> -    sctp_assoc_t            assoc_id;
> -    uint32_t                assoc_value;
> +    sctp_assoc_t    assoc_id;
> +    uint32_t    assoc_value;
> };

Try not to make unrelated changes like whitespace.

> 
> /*
> --- a/net/sctp/socket.c    2008-04-18 09:00:57.000000000 -0400
> +++ b/net/sctp/socket.c    2008-04-19 08:35:21.000000000 -0400
> @@ -2315,69 +2315,91 @@ static int sctp_setsockopt_peer_addr_par
>     return 0;
> }
> 
> -/* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
> +/*
> + * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
> + *
> + * This option will effect the way delayed acks are performed.  This
> + * option allows you to get or set the delayed ack time, in
> + * milliseconds.  It also allows changing the delayed ack frequency.
> + * Changing the frequency to 1 disables the delayed sack algorithm.  If
> + * the assoc_id is 0, then this sets or gets the endpoints default
> + * values.  If the assoc_id field is non-zero, then the set or get
> + * effects the specified association for the one to many model (the
> + * assoc_id field is ignored by the one to one model).  Note that if
> + * sack_delay or sack_freq are 0 when setting this option, then the
> + * current values will remain unchanged.
> + * + * struct sctp_sack_info {
> + *     sctp_assoc_t            sack_assoc_id;
> + *     uint32_t                sack_delay;
> + *     uint32_t                sack_freq;
> + * };
>  *
> - *   This options will get or set the delayed ack timer.  The time is set
> - *   in milliseconds.  If the assoc_id is 0, then this sets or gets the
> - *   endpoints default delayed ack timer value.  If the assoc_id field is
> - *   non-zero, then the set or get effects the specified association.
> - *
> - *   struct sctp_assoc_value {
> - *       sctp_assoc_t            assoc_id;
> - *       uint32_t                assoc_value;
> - *   };
> - *
> - *     assoc_id    - This parameter, indicates which association the
> - *                   user is preforming an action upon. Note that if
> - *                   this field's value is zero then the endpoints
> - *                   default value is changed (effecting future
> - *                   associations only).
> - *
> - *     assoc_value - This parameter contains the number of milliseconds
> - *                   that the user is requesting the delayed ACK timer
> - *                   be set to. Note that this value is defined in
> - *                   the standard to be between 200 and 500 milliseconds.
> - *
> - *                   Note: a value of zero will leave the value alone,
> - *                   but disable SACK delay. A non-zero value will also
> - *                   enable SACK delay.
> + * sack_assoc_id -  This parameter, indicates which association the user
> + *    is performing an action upon.  Note that if this field's value is
> + *    zero then the endpoints default value is changed (effecting future
> + *    associations only).
> + *
> + * sack_delay -  This parameter contains the number of milliseconds that
> + *    the user is requesting the delayed ACK timer be set to.  Note that
> + *    this value is defined in the standard to be between 200 and 500
> + *    milliseconds.
> + *
> + * sack_freq -  This parameter contains the number of packets that must
> + *    be received before a sack is sent without waiting for the delay
> + *    timer to expire.  The default value for this is 2, setting this
> + *    value to 1 will disable the delayed sack algorithm.
>  */
> 
> -static int sctp_setsockopt_delayed_ack_time(struct sock *sk,
> +static int sctp_setsockopt_delayed_ack(struct sock *sk,
>                         char __user *optval, int optlen)
> {
> -    struct sctp_assoc_value  params;
> +    struct sctp_sack_info    params;
>     struct sctp_transport   *trans = NULL;
>     struct sctp_association *asoc = NULL;
>     struct sctp_sock        *sp = sctp_sk(sk);
> 
> -    if (optlen != sizeof(struct sctp_assoc_value))
> -        return - EINVAL;
> +    if (optlen >= sizeof(struct sctp_sack_info)) {

Why >= on a setsockopt?  The new structure is bigger then the old one.
You should test for equality here.

> +        optlen = sizeof(struct sctp_sack_info);
> 
> -    if (copy_from_user(&params, optval, optlen))
> -        return -EFAULT;
> +        if (copy_from_user(&params, optval, optlen))
> +            return -EFAULT;
> +
> +        if (params.sack_delay = 0 && params.sack_freq = 0)
> +            return 0;
> +
> +        if (params.sack_freq)
> +            params.sack_delay = 0;

This is not right.  What if someones sets frequency to 3 and delay to 300 ms.
Those are valid values, but you end up ignoring the different delay.

Additionally, the rest of the patch doesn't seem to implement the sack_freq completely.

> +    } else if (optlen = sizeof(struct sctp_assoc_value)) {
> +        printk(KERN_WARNING "SCTP: Use of struct sctp_sack_info "
> +               "in delayed_ack socket option deprecated\n");
> +        printk(KERN_WARNING "SCTP: struct sctp_sack_info instead\n");
> +        if (copy_from_user(&params, optval, optlen))
> +            return -EFAULT;

Since the params are an on-stack variable, it may contain garbage and the
above copy_from_user will leave that garbage in the sack_freq field. 

> +    } else
> +        return - EINVAL;
> 
>     /* Validate value parameter. */
> -    if (params.assoc_value > 500)
> +    if (params.sack_delay > 500)
>         return -EINVAL;
> 
> -    /* Get association, if assoc_id != 0 and the socket is a one
> +    /* Get association, if sack_assoc_id != 0 and the socket is a one
>      * to many style socket, and an association was not found, then
>      * the id was invalid.
>      */
> -    asoc = sctp_id2assoc(sk, params.assoc_id);
> -    if (!asoc && params.assoc_id && sctp_style(sk, UDP))
> +    asoc = sctp_id2assoc(sk, params.sack_assoc_id);
> +    if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
>         return -EINVAL;
> 
> -    if (params.assoc_value) {
> +    if (params.sack_delay) {
>         if (asoc) {
>             asoc->sackdelay > -                msecs_to_jiffies(params.assoc_value);
> +                msecs_to_jiffies(params.sack_delay);
>             asoc->param_flags >                 (asoc->param_flags & ~SPP_SACKDELAY) |
>                 SPP_SACKDELAY_ENABLE;
>         } else {
> -            sp->sackdelay = params.assoc_value;
> +            sp->sackdelay = params.sack_delay;
>             sp->param_flags >                 (sp->param_flags & ~SPP_SACKDELAY) |
>                 SPP_SACKDELAY_ENABLE;
> @@ -2401,9 +2423,9 @@ static int sctp_setsockopt_delayed_ack_t
>         list_for_each(pos, &asoc->peer.transport_addr_list) {
>             trans = list_entry(pos, struct sctp_transport,
>                        transports);
> -            if (params.assoc_value) {
> +            if (params.sack_delay) {
>                 trans->sackdelay > -                    msecs_to_jiffies(params.assoc_value);
> +                    msecs_to_jiffies(params.sack_delay);
>                 trans->param_flags >                     (trans->param_flags & ~SPP_SACKDELAY) |
>                     SPP_SACKDELAY_ENABLE;
> @@ -3204,8 +3226,8 @@ SCTP_STATIC int sctp_setsockopt(struct s
>         retval = sctp_setsockopt_peer_addr_params(sk, optval, optlen);
>         break;
> 
> -    case SCTP_DELAYED_ACK_TIME:
> -        retval = sctp_setsockopt_delayed_ack_time(sk, optval, optlen);
> +    case SCTP_DELAYED_ACK:
> +        retval = sctp_setsockopt_delayed_ack(sk, optval, optlen);
>         break;

This does not implement the sack_freq setting.

>     case SCTP_PARTIAL_DELIVERY_POINT:
>         retval = sctp_setsockopt_partial_delivery_point(sk, optval, 
> optlen);
> @@ -4017,70 +4039,90 @@ static int sctp_getsockopt_peer_addr_par
>     return 0;
> }
> 
> -/* 7.1.23. Delayed Ack Timer (SCTP_DELAYED_ACK_TIME)
> +/*
> + * 7.1.23.  Get or set delayed ack timer (SCTP_DELAYED_SACK)
>  *
> - *   This options will get or set the delayed ack timer.  The time is set
> - *   in milliseconds.  If the assoc_id is 0, then this sets or gets the
> - *   endpoints default delayed ack timer value.  If the assoc_id field is
> - *   non-zero, then the set or get effects the specified association.
> - *
> - *   struct sctp_assoc_value {
> - *       sctp_assoc_t            assoc_id;
> - *       uint32_t                assoc_value;
> - *   };
> - *
> - *     assoc_id    - This parameter, indicates which association the
> - *                   user is preforming an action upon. Note that if
> - *                   this field's value is zero then the endpoints
> - *                   default value is changed (effecting future
> - *                   associations only).
> - *
> - *     assoc_value - This parameter contains the number of milliseconds
> - *                   that the user is requesting the delayed ACK timer
> - *                   be set to. Note that this value is defined in
> - *                   the standard to be between 200 and 500 milliseconds.
> - *
> - *                   Note: a value of zero will leave the value alone,
> - *                   but disable SACK delay. A non-zero value will also
> - *                   enable SACK delay.
> + * This option will effect the way delayed acks are performed.  This
> + * option allows you to get or set the delayed ack time, in
> + * milliseconds.  It also allows changing the delayed ack frequency.
> + * Changing the frequency to 1 disables the delayed sack algorithm.  If
> + * the assoc_id is 0, then this sets or gets the endpoints default
> + * values.  If the assoc_id field is non-zero, then the set or get
> + * effects the specified association for the one to many model (the
> + * assoc_id field is ignored by the one to one model).  Note that if
> + * sack_delay or sack_freq are 0 when setting this option, then the
> + * current values will remain unchanged.
> + * + * struct sctp_sack_info {
> + *     sctp_assoc_t            sack_assoc_id;
> + *     uint32_t                sack_delay;
> + *     uint32_t                sack_freq;
> + * };
> + *
> + * sack_assoc_id -  This parameter, indicates which association the user
> + *    is performing an action upon.  Note that if this field's value is
> + *    zero then the endpoints default value is changed (effecting future
> + *    associations only).
> + *
> + * sack_delay -  This parameter contains the number of milliseconds that
> + *    the user is requesting the delayed ACK timer be set to.  Note that
> + *    this value is defined in the standard to be between 200 and 500
> + *    milliseconds.
> + *
> + * sack_freq -  This parameter contains the number of packets that must
> + *    be received before a sack is sent without waiting for the delay
> + *    timer to expire.  The default value for this is 2, setting this
> + *    value to 1 will disable the delayed sack algorithm.
>  */
> -static int sctp_getsockopt_delayed_ack_time(struct sock *sk, int len,
> +static int sctp_getsockopt_delayed_ack(struct sock *sk, int len,
>                         char __user *optval,
>                         int __user *optlen)
> {
> -    struct sctp_assoc_value  params;
> +    struct sctp_sack_info    params;
>     struct sctp_association *asoc = NULL;
>     struct sctp_sock        *sp = sctp_sk(sk);
> 
> -    if (len < sizeof(struct sctp_assoc_value))
> -        return - EINVAL;
> -
> -    len = sizeof(struct sctp_assoc_value);
> +    if (len >= sizeof(struct sctp_sack_info)) {
> +        len = sizeof(struct sctp_sack_info);

This has a potential to break applications.

> 
> -    if (copy_from_user(&params, optval, len))
> -        return -EFAULT;
> +        if (copy_from_user(&params, optval, len))
> +            return -EFAULT;
> +    } else if (len = sizeof(struct sctp_assoc_value)) {
> +        printk(KERN_WARNING "SCTP: Use of struct sctp_sack_info "
> +               "in delayed_ack socket option deprecated\n");
> +        printk(KERN_WARNING "SCTP: struct sctp_sack_info instead\n");
> +        if (copy_from_user(&params, optval, len))
> +            return -EFAULT;
> +    } else +        return - EINVAL;

New line please.
Also, you need to zero out params for the same reason as above.

> 
> -    /* Get association, if assoc_id != 0 and the socket is a one
> +    /* Get association, if sack_assoc_id != 0 and the socket is a one
>      * to many style socket, and an association was not found, then
>      * the id was invalid.
>      */
> -    asoc = sctp_id2assoc(sk, params.assoc_id);
> -    if (!asoc && params.assoc_id && sctp_style(sk, UDP))
> +    asoc = sctp_id2assoc(sk, params.sack_assoc_id);
> +    if (!asoc && params.sack_assoc_id && sctp_style(sk, UDP))
>         return -EINVAL;
> 
>     if (asoc) {
>         /* Fetch association values. */
> -        if (asoc->param_flags & SPP_SACKDELAY_ENABLE)
> -            params.assoc_value = jiffies_to_msecs(
> +        if (asoc->param_flags & SPP_SACKDELAY_ENABLE) {
> +            params.sack_delay = jiffies_to_msecs(
>                 asoc->sackdelay);
> -        else
> -            params.assoc_value = 0;
> +            params.sack_freq = 0;

sack_freq is 2 by default, but this patch should implement the ability to set it.

> +        } else {
> +            params.sack_delay = 0;
> +            params.sack_freq = 1;
> +        }
>     } else {
>         /* Fetch socket values. */
> -        if (sp->param_flags & SPP_SACKDELAY_ENABLE)
> -            params.assoc_value  = sp->sackdelay;
> -        else
> -            params.assoc_value  = 0;
> +        if (sp->param_flags & SPP_SACKDELAY_ENABLE) {
> +            params.sack_delay  = sp->sackdelay;
> +            params.sack_freq = 0;

Same as above.

> +        } else {
> +            params.sack_delay  = 0;
> +            params.sack_freq = 1;
> +        }
>     }
> 
>     if (copy_to_user(optval, &params, len))
> @@ -5238,8 +5280,8 @@ SCTP_STATIC int sctp_getsockopt(struct s
>         retval = sctp_getsockopt_peer_addr_params(sk, len, optval,
>                               optlen);
>         break;
> -    case SCTP_DELAYED_ACK_TIME:
> -        retval = sctp_getsockopt_delayed_ack_time(sk, len, optval,
> +    case SCTP_DELAYED_ACK:
> +        retval = sctp_getsockopt_delayed_ack(sk, len, optval,
>                               optlen);
>         break;

-vlad