[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <87wo94asxt.fsf@mellanox.com>
Date: Mon, 03 Feb 2020 11:03:42 +0100
From: Petr Machata <petrm@...lanox.com>
To: Nathan Chancellor <natechancellor@...il.com>
Cc: Randy Dunlap <rdunlap@...radead.org>,
Jiri Pirko <jiri@...lanox.com>,
Ido Schimmel <idosch@...lanox.com>,
"David S. Miller" <davem@...emloft.net>, netdev@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH] mlxsw: spectrum_qdisc: Fix 64-bit division error in mlxsw_sp_qdisc_tbf_rate_kbps
Nathan Chancellor <natechancellor@...il.com> writes:
> On Thu, Jan 30, 2020 at 05:43:56PM -0800, Randy Dunlap wrote:
>> On 1/30/20 3:26 PM, Nathan Chancellor wrote:
>> > When building arm32 allmodconfig:
>> >
>> > ERROR: "__aeabi_uldivmod"
>> > [drivers/net/ethernet/mellanox/mlxsw/mlxsw_spectrum.ko] undefined!
>> >
>> > rate_bytes_ps has type u64, we need to use a 64-bit division helper to
>> > avoid a build error.
>> >
>> > Fixes: a44f58c41bfb ("mlxsw: spectrum_qdisc: Support offloading of TBF Qdisc")
>> > Signed-off-by: Nathan Chancellor <natechancellor@...il.com>
>> > ---
>> > drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c | 2 +-
>> > 1 file changed, 1 insertion(+), 1 deletion(-)
>> >
>> > diff --git a/drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c b/drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c
>> > index 79a2801d59f6..65e681ef01e8 100644
>> > --- a/drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c
>> > +++ b/drivers/net/ethernet/mellanox/mlxsw/spectrum_qdisc.c
>> > @@ -614,7 +614,7 @@ mlxsw_sp_qdisc_tbf_rate_kbps(struct tc_tbf_qopt_offload_replace_params *p)
>> > /* TBF interface is in bytes/s, whereas Spectrum ASIC is configured in
>> > * Kbits/s.
>> > */
>> > - return p->rate.rate_bytes_ps / 1000 * 8;
>> > + return div_u64(p->rate.rate_bytes_ps, 1000 * 8);
>>
>> not quite right AFAICT.
>>
>> try either
>> return div_u64(p->rate.rate_bytes_ps * 8, 1000);
>> or
>> return div_u64(p->rate.rate_bytes_ps, 1000) * 8;
>>
>
> Gah, I swear I can math... Thank you for catching this, v2 incoming with
> the later because I think it looks better.
Yes, that's the correct choice. Divide first, that way we can't
overflow.
Thanks for taking care of this.
Powered by blists - more mailing lists