On some hardware units it was found this trace is flooding the output, making any dpdk interactive usage kind of problematic. It is only informational, without any consequences handling, so reducing it to verbose from explicit notice level. CC: stable@dpdk.org Signed-off-by: Igor Russkikh <irusskikh@marvell.com> --- drivers/net/qede/base/ecore_int.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/net/qede/base/ecore_int.c b/drivers/net/qede/base/ecore_int.c index 4207b1853..2c4aac941 100644 --- a/drivers/net/qede/base/ecore_int.c +++ b/drivers/net/qede/base/ecore_int.c @@ -928,7 +928,7 @@ static void ecore_int_attn_print(struct ecore_hwfn *p_hwfn, bool b_clear) { /* @DPDK */ - DP_NOTICE(p_hwfn->p_dev, false, "[block_id %d type %d]\n", id, type); + DP_VERBOSE(p_hwfn, ECORE_MSG_INTR, "[block_id %d type %d]\n", id, type); } /** -- 2.25.1
We found some dpdk applications blindly pass fixed side rss hash tables, and do not check driver/device capabilities. Moreover, many other drivers do not do such a strong check as well. So here we fix it, making qede accept any size rss_key. For larger key tables we just crop it with notice trace message. CC: stable@dpdk.org Signed-off-by: Igor Russkikh <irusskikh@marvell.com> --- drivers/net/qede/qede_ethdev.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/drivers/net/qede/qede_ethdev.c b/drivers/net/qede/qede_ethdev.c index ab5f5b106..7363d98f2 100644 --- a/drivers/net/qede/qede_ethdev.c +++ b/drivers/net/qede/qede_ethdev.c @@ -2139,8 +2139,8 @@ int qede_rss_hash_update(struct rte_eth_dev *eth_dev, /* RSS hash key */ if (key) { if (len > (ECORE_RSS_KEY_SIZE * sizeof(uint32_t))) { - DP_ERR(edev, "RSS key length exceeds limit\n"); - return -EINVAL; + len = ECORE_RSS_KEY_SIZE * sizeof(uint32_t); + DP_NOTICE(edev, false, "RSS key length exceeds limit\n"); } DP_INFO(edev, "Applying user supplied hash key\n"); rss_params.update_rss_key = 1; -- 2.25.1
On Sun, Feb 21, 2021 at 4:47 PM Igor Russkikh <irusskikh@marvell.com> wrote: > > We found some dpdk applications blindly pass fixed side rss hash tables, > and do not check driver/device capabilities. > > Moreover, many other drivers do not do such a strong check as well. > > So here we fix it, making qede accept any size rss_key. For larger key > tables we just crop it with notice trace message. > > CC: stable@dpdk.org > Signed-off-by: Igor Russkikh <irusskikh@marvell.com> HI @Rasesh Mody @Shahed Shaikh Could you review this series? > --- > drivers/net/qede/qede_ethdev.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/net/qede/qede_ethdev.c b/drivers/net/qede/qede_ethdev.c > index ab5f5b106..7363d98f2 100644 > --- a/drivers/net/qede/qede_ethdev.c > +++ b/drivers/net/qede/qede_ethdev.c > @@ -2139,8 +2139,8 @@ int qede_rss_hash_update(struct rte_eth_dev *eth_dev, > /* RSS hash key */ > if (key) { > if (len > (ECORE_RSS_KEY_SIZE * sizeof(uint32_t))) { > - DP_ERR(edev, "RSS key length exceeds limit\n"); > - return -EINVAL; > + len = ECORE_RSS_KEY_SIZE * sizeof(uint32_t); > + DP_NOTICE(edev, false, "RSS key length exceeds limit\n"); > } > DP_INFO(edev, "Applying user supplied hash key\n"); > rss_params.update_rss_key = 1; > -- > 2.25.1 >
> -----Original Message-----
> From: Igor Russkikh <irusskikh@marvell.com>
> Sent: Sunday, February 21, 2021 4:47 PM
> To: dev@dpdk.org
> Cc: Rasesh Mody <rmody@marvell.com>; Devendra Singh Rawat
> <dsinghrawat@marvell.com>; Igor Russkikh <irusskikh@marvell.com>;
> stable@dpdk.org
> Subject: [PATCH 1/2] qede: reduce trace verbosity level
>
> On some hardware units it was found this trace is flooding the output, making
> any dpdk interactive usage kind of problematic.
>
> It is only informational, without any consequences handling, so reducing it to
> verbose from explicit notice level.
>
> CC: stable@dpdk.org
> Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
> ---
> drivers/net/qede/base/ecore_int.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/net/qede/base/ecore_int.c
> b/drivers/net/qede/base/ecore_int.c
> index 4207b1853..2c4aac941 100644
> --- a/drivers/net/qede/base/ecore_int.c
> +++ b/drivers/net/qede/base/ecore_int.c
> @@ -928,7 +928,7 @@ static void ecore_int_attn_print(struct ecore_hwfn
> *p_hwfn,
> bool b_clear)
> {
> /* @DPDK */
> - DP_NOTICE(p_hwfn->p_dev, false, "[block_id %d type %d]\n", id,
> type);
> + DP_VERBOSE(p_hwfn, ECORE_MSG_INTR, "[block_id %d type %d]\n",
> id,
> +type);
> }
>
> /**
> --
> 2.25.1
ACK,
Acked-by: Devendra Singh Rawat <dsinghrawat@marvell.com>
Thanks,
Devendra
> -----Original Message-----
> From: Igor Russkikh <irusskikh@marvell.com>
> Sent: Sunday, February 21, 2021 4:47 PM
> To: dev@dpdk.org
> Cc: Rasesh Mody <rmody@marvell.com>; Devendra Singh Rawat
> <dsinghrawat@marvell.com>; Igor Russkikh <irusskikh@marvell.com>;
> stable@dpdk.org
> Subject: [PATCH 2/2] qede: make driver accept bigger rss tables
>
> We found some dpdk applications blindly pass fixed side rss hash tables, and
> do not check driver/device capabilities.
>
> Moreover, many other drivers do not do such a strong check as well.
>
> So here we fix it, making qede accept any size rss_key. For larger key tables we
> just crop it with notice trace message.
>
> CC: stable@dpdk.org
> Signed-off-by: Igor Russkikh <irusskikh@marvell.com>
> ---
> drivers/net/qede/qede_ethdev.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/qede/qede_ethdev.c b/drivers/net/qede/qede_ethdev.c
> index ab5f5b106..7363d98f2 100644
> --- a/drivers/net/qede/qede_ethdev.c
> +++ b/drivers/net/qede/qede_ethdev.c
> @@ -2139,8 +2139,8 @@ int qede_rss_hash_update(struct rte_eth_dev
> *eth_dev,
> /* RSS hash key */
> if (key) {
> if (len > (ECORE_RSS_KEY_SIZE * sizeof(uint32_t))) {
> - DP_ERR(edev, "RSS key length exceeds
> limit\n");
> - return -EINVAL;
> + len = ECORE_RSS_KEY_SIZE * sizeof(uint32_t);
> + DP_NOTICE(edev, false, "RSS key length
> exceeds limit\n");
IMO, it will be better if the log message also states that RSS key is trimmed to 'len' size.
Thanks,
Devendra
> IMO, it will be better if the log message also states that RSS key is trimmed to 'len' size.
Resent v2, thanks.
Devendra, please take these two patches into our internal codebase as well.
Igor
Sure Igor, I will do that.
Thanks,
Devendra
> -----Original Message-----
> From: Igor Russkikh <irusskikh@marvell.com>
> Sent: Friday, March 19, 2021 3:18 PM
> To: Devendra Singh Rawat <dsinghrawat@marvell.com>; dev@dpdk.org
> Cc: Rasesh Mody <rmody@marvell.com>; stable@dpdk.org
> Subject: RE: [PATCH 2/2] qede: make driver accept bigger rss tables
>
>
> > IMO, it will be better if the log message also states that RSS key is trimmed to
> 'len' size.
>
> Resent v2, thanks.
>
> Devendra, please take these two patches into our internal codebase as well.
>
> Igor