| Message ID | 20230810002313.421684-1-jesse.brandeburg@intel.com (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [iwl-net,v1] ice: fix receive buffer size miscalculation | expand |
On 8/10/23 02:23, Jesse Brandeburg wrote: > The driver is misconfiguring the hardware for some values of MTU such that > it could use multiple descriptors to receive a packet when it could have > simply used one. > > Change the driver to use a round-up instead of the result of a shift, as > the shift can truncate the lower bits of the size, and result in the > problem noted above. It also aligns this driver with similar code in i40e. > > The insidiousness of this problem is that everything works with the wrong > size, it's just not working as well as it could, as some MTU sizes end up > using two or more descriptors, and there is no way to tell that is > happening without looking at ice_trace or a bus analyzer. > > Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com> Nice catch! > --- > drivers/net/ethernet/intel/ice/ice_base.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c > index b678bdf96f3a..074bf9403cd1 100644 > --- a/drivers/net/ethernet/intel/ice/ice_base.c > +++ b/drivers/net/ethernet/intel/ice/ice_base.c > @@ -435,7 +435,8 @@ static int ice_setup_rx_ctx(struct ice_rx_ring *ring) > /* Receive Packet Data Buffer Size. > * The Packet Data Buffer Size is defined in 128 byte units. > */ > - rlan_ctx.dbuf = ring->rx_buf_len >> ICE_RLAN_CTX_DBUF_S; > + rlan_ctx.dbuf = DIV_ROUND_UP(ring->rx_buf_len, > + BIT_ULL(ICE_RLAN_CTX_DBUF_S)); looks like %ICE_RLAN_CTX_DBUF_S is only used once in whole codebase, here. Perhaps you could rename it (even considering that it is targeted for -net), and have it as ICE_RLAN_CTX_DBUF_DIV or similar? Then it will be BIT(7) or just 128. Also, I would keep it as 32bit value. > > /* use 32 byte descriptors */ > rlan_ctx.dsize = 1; (BTW: you have missed my review on e1000 ML) anyway, it's good Reviewed-by: Przemek Kitszel <przemyslaw.kitszel@intel.com>
On 8/9/2023 5:23 PM, Jesse Brandeburg wrote: > The driver is misconfiguring the hardware for some values of MTU such that > it could use multiple descriptors to receive a packet when it could have > simply used one. > > Change the driver to use a round-up instead of the result of a shift, as > the shift can truncate the lower bits of the size, and result in the > problem noted above. It also aligns this driver with similar code in i40e. > > The insidiousness of this problem is that everything works with the wrong > size, it's just not working as well as it could, as some MTU sizes end up > using two or more descriptors, and there is no way to tell that is > happening without looking at ice_trace or a bus analyzer. This should have a Fixes: ? > Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com> > --- > drivers/net/ethernet/intel/ice/ice_base.c | 3 ++- > 1 file changed, 2 insertions(+), 1 deletion(-) > > diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c > index b678bdf96f3a..074bf9403cd1 100644 > --- a/drivers/net/ethernet/intel/ice/ice_base.c > +++ b/drivers/net/ethernet/intel/ice/ice_base.c > @@ -435,7 +435,8 @@ static int ice_setup_rx_ctx(struct ice_rx_ring *ring) > /* Receive Packet Data Buffer Size. > * The Packet Data Buffer Size is defined in 128 byte units. > */ > - rlan_ctx.dbuf = ring->rx_buf_len >> ICE_RLAN_CTX_DBUF_S; > + rlan_ctx.dbuf = DIV_ROUND_UP(ring->rx_buf_len, > + BIT_ULL(ICE_RLAN_CTX_DBUF_S)); > > /* use 32 byte descriptors */ > rlan_ctx.dsize = 1;
On 8/10/2023 2:33 PM, Tony Nguyen wrote: > > > On 8/9/2023 5:23 PM, Jesse Brandeburg wrote: >> The driver is misconfiguring the hardware for some values of MTU such >> that >> it could use multiple descriptors to receive a packet when it could have >> simply used one. >> >> Change the driver to use a round-up instead of the result of a shift, as >> the shift can truncate the lower bits of the size, and result in the >> problem noted above. It also aligns this driver with similar code in >> i40e. >> >> The insidiousness of this problem is that everything works with the wrong >> size, it's just not working as well as it could, as some MTU sizes end up >> using two or more descriptors, and there is no way to tell that is >> happening without looking at ice_trace or a bus analyzer. > > This should have a Fixes: ? Dang, you're correct. I'll send a v2. Thanks, Jesse
diff --git a/drivers/net/ethernet/intel/ice/ice_base.c b/drivers/net/ethernet/intel/ice/ice_base.c index b678bdf96f3a..074bf9403cd1 100644 --- a/drivers/net/ethernet/intel/ice/ice_base.c +++ b/drivers/net/ethernet/intel/ice/ice_base.c @@ -435,7 +435,8 @@ static int ice_setup_rx_ctx(struct ice_rx_ring *ring) /* Receive Packet Data Buffer Size. * The Packet Data Buffer Size is defined in 128 byte units. */ - rlan_ctx.dbuf = ring->rx_buf_len >> ICE_RLAN_CTX_DBUF_S; + rlan_ctx.dbuf = DIV_ROUND_UP(ring->rx_buf_len, + BIT_ULL(ICE_RLAN_CTX_DBUF_S)); /* use 32 byte descriptors */ rlan_ctx.dsize = 1;
The driver is misconfiguring the hardware for some values of MTU such that it could use multiple descriptors to receive a packet when it could have simply used one. Change the driver to use a round-up instead of the result of a shift, as the shift can truncate the lower bits of the size, and result in the problem noted above. It also aligns this driver with similar code in i40e. The insidiousness of this problem is that everything works with the wrong size, it's just not working as well as it could, as some MTU sizes end up using two or more descriptors, and there is no way to tell that is happening without looking at ice_trace or a bus analyzer. Signed-off-by: Jesse Brandeburg <jesse.brandeburg@intel.com> --- drivers/net/ethernet/intel/ice/ice_base.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)