Message ID | 20180113113445.2705-1-jcd@tribudubois.net (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
On 13 January 2018 at 11:34, Jean-Christophe Dubois <jcd@tribudubois.net> wrote: > The actual imx_eth_enable_rx() function is buggy. > > It updates s->regs[ENET_RDAR] after calling qemu_flush_queued_packets(). > > qemu_flush_queued_packets() is going to call imx_XXX_receive() which itself > is going to call imx_eth_enable_rx(). > > By updating s->regs[ENET_RDAR] after calling qemu_flush_queued_packets() > we end up updating the register with an outdated value which might > lead to disabling the receive function in the i.MX FEC/ENET device. > > This patch change the place where the register update is done so that the > register value stays up to date and the receive function can keep > running. > > Reported-by: Fyleo <fyleo45@gmail.com> > Tested-by: Fyleo <fyleo45@gmail.com> > Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net> > --- Andrey, do you have an opinion on this patch, since you've been looking at i.MX code recently? > Change since v1: > 1. Rebase the patch on the updated master branch > > hw/net/imx_fec.c | 8 ++------ > 1 file changed, 2 insertions(+), 6 deletions(-) > > diff --git a/hw/net/imx_fec.c b/hw/net/imx_fec.c > index 4fb48f62ba..9506f9b69f 100644 > --- a/hw/net/imx_fec.c > +++ b/hw/net/imx_fec.c > @@ -595,19 +595,16 @@ static void imx_eth_do_tx(IMXFECState *s, uint32_t index) > static void imx_eth_enable_rx(IMXFECState *s, bool flush) > { > IMXFECBufDesc bd; > - bool rx_ring_full; > > imx_fec_read_bd(&bd, s->rx_descriptor); > > - rx_ring_full = !(bd.flags & ENET_BD_E); > + s->regs[ENET_RDAR] = (bd.flags & ENET_BD_E) ? ENET_RDAR_RDAR : 0; > > - if (rx_ring_full) { > + if (!s->regs[ENET_RDAR]) { > FEC_PRINTF("RX buffer full\n"); > } else if (flush) { > qemu_flush_queued_packets(qemu_get_queue(s->nic)); > } > - > - s->regs[ENET_RDAR] = rx_ring_full ? 0 : ENET_RDAR_RDAR; > } > > static void imx_eth_reset(DeviceState *d) > @@ -866,7 +863,6 @@ static void imx_eth_write(void *opaque, hwaddr offset, uint64_t value, > case ENET_RDAR: > if (s->regs[ENET_ECR] & ENET_ECR_ETHEREN) { > if (!s->regs[index]) { > - s->regs[index] = ENET_RDAR_RDAR; > imx_eth_enable_rx(s, true); > } > } else { > -- > 2.14.1 Reviewed-by: Peter Maydell <peter.maydell@linaro.org> thanks -- PMM
On Mon, Jan 22, 2018 at 3:48 AM, Peter Maydell <peter.maydell@linaro.org> wrote: > On 13 January 2018 at 11:34, Jean-Christophe Dubois <jcd@tribudubois.net> wrote: >> The actual imx_eth_enable_rx() function is buggy. >> >> It updates s->regs[ENET_RDAR] after calling qemu_flush_queued_packets(). >> >> qemu_flush_queued_packets() is going to call imx_XXX_receive() which itself >> is going to call imx_eth_enable_rx(). >> >> By updating s->regs[ENET_RDAR] after calling qemu_flush_queued_packets() >> we end up updating the register with an outdated value which might >> lead to disabling the receive function in the i.MX FEC/ENET device. >> >> This patch change the place where the register update is done so that the >> register value stays up to date and the receive function can keep >> running. >> >> Reported-by: Fyleo <fyleo45@gmail.com> >> Tested-by: Fyleo <fyleo45@gmail.com> >> Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net> >> --- > > Andrey, do you have an opinion on this patch, since you've been > looking at i.MX code recently? > The rationale makes sense to me and patch looks like a good cleanup in general, so FWIW: Reviewed-by: Andrey Smirnov <andrew.smirnov@gmail.com> I also gave it a spin against my i.MX7 changes with doing basic things like ping and scp of 1GB file, so I can give my: Tested-by: Andrey Smirnov <andrew.smirnov@gmail.com> for that. Thanks, Andrey Smirnov >> Change since v1: >> 1. Rebase the patch on the updated master branch >> >> hw/net/imx_fec.c | 8 ++------ >> 1 file changed, 2 insertions(+), 6 deletions(-) >> >> diff --git a/hw/net/imx_fec.c b/hw/net/imx_fec.c >> index 4fb48f62ba..9506f9b69f 100644 >> --- a/hw/net/imx_fec.c >> +++ b/hw/net/imx_fec.c >> @@ -595,19 +595,16 @@ static void imx_eth_do_tx(IMXFECState *s, uint32_t index) >> static void imx_eth_enable_rx(IMXFECState *s, bool flush) >> { >> IMXFECBufDesc bd; >> - bool rx_ring_full; >> >> imx_fec_read_bd(&bd, s->rx_descriptor); >> >> - rx_ring_full = !(bd.flags & ENET_BD_E); >> + s->regs[ENET_RDAR] = (bd.flags & ENET_BD_E) ? ENET_RDAR_RDAR : 0; >> >> - if (rx_ring_full) { >> + if (!s->regs[ENET_RDAR]) { >> FEC_PRINTF("RX buffer full\n"); >> } else if (flush) { >> qemu_flush_queued_packets(qemu_get_queue(s->nic)); >> } >> - >> - s->regs[ENET_RDAR] = rx_ring_full ? 0 : ENET_RDAR_RDAR; >> } >> >> static void imx_eth_reset(DeviceState *d) >> @@ -866,7 +863,6 @@ static void imx_eth_write(void *opaque, hwaddr offset, uint64_t value, >> case ENET_RDAR: >> if (s->regs[ENET_ECR] & ENET_ECR_ETHEREN) { >> if (!s->regs[index]) { >> - s->regs[index] = ENET_RDAR_RDAR; >> imx_eth_enable_rx(s, true); >> } >> } else { >> -- >> 2.14.1 > > Reviewed-by: Peter Maydell <peter.maydell@linaro.org> > > thanks > -- PMM
On 22 January 2018 at 14:54, Andrey Smirnov <andrew.smirnov@gmail.com> wrote: > On Mon, Jan 22, 2018 at 3:48 AM, Peter Maydell <peter.maydell@linaro.org> wrote: >> On 13 January 2018 at 11:34, Jean-Christophe Dubois <jcd@tribudubois.net> wrote: >>> The actual imx_eth_enable_rx() function is buggy. >>> >>> It updates s->regs[ENET_RDAR] after calling qemu_flush_queued_packets(). >>> >>> qemu_flush_queued_packets() is going to call imx_XXX_receive() which itself >>> is going to call imx_eth_enable_rx(). >>> >>> By updating s->regs[ENET_RDAR] after calling qemu_flush_queued_packets() >>> we end up updating the register with an outdated value which might >>> lead to disabling the receive function in the i.MX FEC/ENET device. >>> >>> This patch change the place where the register update is done so that the >>> register value stays up to date and the receive function can keep >>> running. >>> >>> Reported-by: Fyleo <fyleo45@gmail.com> >>> Tested-by: Fyleo <fyleo45@gmail.com> >>> Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net> >>> --- >> >> Andrey, do you have an opinion on this patch, since you've been >> looking at i.MX code recently? >> > > The rationale makes sense to me and patch looks like a good cleanup in > general, so FWIW: > > Reviewed-by: Andrey Smirnov <andrew.smirnov@gmail.com> > > I also gave it a spin against my i.MX7 changes with doing basic things > like ping and scp of 1GB file, so I can give my: > > Tested-by: Andrey Smirnov <andrew.smirnov@gmail.com> Thanks; I've applied the patch to target-arm.next. -- PMM
My guess is that with this patch, the "flush" feature that was added by Andrey in "imx_fec: Change queue flushing heuristics" (commit b2b012a) is not really necessary anymore. But it does not hurt (it might induce a little bit more processing). JC Le 22/01/2018 à 15:59, Peter Maydell a écrit : > On 22 January 2018 at 14:54, Andrey Smirnov <andrew.smirnov@gmail.com> wrote: >> On Mon, Jan 22, 2018 at 3:48 AM, Peter Maydell <peter.maydell@linaro.org> wrote: >>> On 13 January 2018 at 11:34, Jean-Christophe Dubois <jcd@tribudubois.net> wrote: >>>> The actual imx_eth_enable_rx() function is buggy. >>>> >>>> It updates s->regs[ENET_RDAR] after calling qemu_flush_queued_packets(). >>>> >>>> qemu_flush_queued_packets() is going to call imx_XXX_receive() which itself >>>> is going to call imx_eth_enable_rx(). >>>> >>>> By updating s->regs[ENET_RDAR] after calling qemu_flush_queued_packets() >>>> we end up updating the register with an outdated value which might >>>> lead to disabling the receive function in the i.MX FEC/ENET device. >>>> >>>> This patch change the place where the register update is done so that the >>>> register value stays up to date and the receive function can keep >>>> running. >>>> >>>> Reported-by: Fyleo <fyleo45@gmail.com> >>>> Tested-by: Fyleo <fyleo45@gmail.com> >>>> Signed-off-by: Jean-Christophe Dubois <jcd@tribudubois.net> >>>> --- >>> Andrey, do you have an opinion on this patch, since you've been >>> looking at i.MX code recently? >>> >> The rationale makes sense to me and patch looks like a good cleanup in >> general, so FWIW: >> >> Reviewed-by: Andrey Smirnov <andrew.smirnov@gmail.com> >> >> I also gave it a spin against my i.MX7 changes with doing basic things >> like ping and scp of 1GB file, so I can give my: >> >> Tested-by: Andrey Smirnov <andrew.smirnov@gmail.com> > Thanks; I've applied the patch to target-arm.next. > > -- PMM > >
diff --git a/hw/net/imx_fec.c b/hw/net/imx_fec.c index 4fb48f62ba..9506f9b69f 100644 --- a/hw/net/imx_fec.c +++ b/hw/net/imx_fec.c @@ -595,19 +595,16 @@ static void imx_eth_do_tx(IMXFECState *s, uint32_t index) static void imx_eth_enable_rx(IMXFECState *s, bool flush) { IMXFECBufDesc bd; - bool rx_ring_full; imx_fec_read_bd(&bd, s->rx_descriptor); - rx_ring_full = !(bd.flags & ENET_BD_E); + s->regs[ENET_RDAR] = (bd.flags & ENET_BD_E) ? ENET_RDAR_RDAR : 0; - if (rx_ring_full) { + if (!s->regs[ENET_RDAR]) { FEC_PRINTF("RX buffer full\n"); } else if (flush) { qemu_flush_queued_packets(qemu_get_queue(s->nic)); } - - s->regs[ENET_RDAR] = rx_ring_full ? 0 : ENET_RDAR_RDAR; } static void imx_eth_reset(DeviceState *d) @@ -866,7 +863,6 @@ static void imx_eth_write(void *opaque, hwaddr offset, uint64_t value, case ENET_RDAR: if (s->regs[ENET_ECR] & ENET_ECR_ETHEREN) { if (!s->regs[index]) { - s->regs[index] = ENET_RDAR_RDAR; imx_eth_enable_rx(s, true); } } else {