[PATCH 1/4] ALSA: emu10k1: simplify interrupt handler, part 1
IPR_CHANNELNUMBERMASK cannot be non-zero when IPR_CHANNELLOOP is unset, so join marking them as handled. This logically reverts part of commit f453e20d8a0 ("ALSA update 0.9.3a"), which made the inverse change with no explanation.
Signed-off-by: Oswald Buddenhagen oswald.buddenhagen@gmx.de --- sound/pci/emu10k1/irq.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-)
diff --git a/sound/pci/emu10k1/irq.c b/sound/pci/emu10k1/irq.c index dfb44e5e69a7..0cb89bd8c16b 100644 --- a/sound/pci/emu10k1/irq.c +++ b/sound/pci/emu10k1/irq.c @@ -79,9 +79,8 @@ irqreturn_t snd_emu10k1_interrupt(int irq, void *dev_id) val >>= 1; pvoice++; } - status &= ~IPR_CHANNELLOOP; + status &= ~(IPR_CHANNELLOOP | IPR_CHANNELNUMBERMASK); } - status &= ~IPR_CHANNELNUMBERMASK; if (status & (IPR_ADCBUFFULL|IPR_ADCBUFHALFFULL)) { if (emu->capture_interrupt) emu->capture_interrupt(emu, status);
Remove weird INTE_* clearing code. The bits were a subset of the actually handled interrupts, which kind of contradicted the stated purpose. I suppose it would make sense to complete the set and negate it, but interrupts being enabled out of the blue is neither something that happens a lot, nor should it result in just one error message, IMO.
Signed-off-by: Oswald Buddenhagen oswald.buddenhagen@gmx.de --- sound/pci/emu10k1/irq.c | 18 ------------------ 1 file changed, 18 deletions(-)
diff --git a/sound/pci/emu10k1/irq.c b/sound/pci/emu10k1/irq.c index 0cb89bd8c16b..312511300053 100644 --- a/sound/pci/emu10k1/irq.c +++ b/sound/pci/emu10k1/irq.c @@ -146,26 +146,8 @@ irqreturn_t snd_emu10k1_interrupt(int irq, void *dev_id) }
if (status) { - unsigned int bits; dev_err(emu->card->dev, "unhandled interrupt: 0x%08x\n", status); - //make sure any interrupts we don't handle are disabled: - bits = INTE_FXDSPENABLE | - INTE_PCIERRORENABLE | - INTE_VOLINCRENABLE | - INTE_VOLDECRENABLE | - INTE_MUTEENABLE | - INTE_MICBUFENABLE | - INTE_ADCBUFENABLE | - INTE_EFXBUFENABLE | - INTE_GPSPDIFENABLE | - INTE_CDSPDIFENABLE | - INTE_INTERVALTIMERENB | - INTE_MIDITXENABLE | - INTE_MIDIRXENABLE; - if (emu->audigy) - bits |= INTE_A_MIDITXENABLE2 | INTE_A_MIDIRXENABLE2; - snd_emu10k1_intr_disable(emu, bits); } outl(orig_status, emu->port + IPR); /* ack all */ }
Handle the "timeout" (actually the retry counter) such that it's more obvious and causes less cost in the normal case.
Signed-off-by: Oswald Buddenhagen oswald.buddenhagen@gmx.de --- sound/pci/emu10k1/irq.c | 11 ++++++----- 1 file changed, 6 insertions(+), 5 deletions(-)
diff --git a/sound/pci/emu10k1/irq.c b/sound/pci/emu10k1/irq.c index 312511300053..7dc803aaa850 100644 --- a/sound/pci/emu10k1/irq.c +++ b/sound/pci/emu10k1/irq.c @@ -22,15 +22,18 @@ irqreturn_t snd_emu10k1_interrupt(int irq, void *dev_id) int handled = 0; int timeout = 0;
- while (((status = inl(emu->port + IPR)) != 0) && (timeout < 1000)) { - timeout++; - orig_status = status; + while ((status = inl(emu->port + IPR)) != 0) { handled = 1; if ((status & 0xffffffff) == 0xffffffff) { dev_info(emu->card->dev, "Suspected sound card removal\n"); break; } + if (++timeout == 1000) { + dev_info(emu->card->dev, "emu10k1 irq routine failure\n"); + break; + } + orig_status = status; if (status & IPR_PCIERROR) { dev_err(emu->card->dev, "interrupt: PCI error\n"); snd_emu10k1_intr_disable(emu, INTE_PCIERRORENABLE); @@ -151,8 +154,6 @@ irqreturn_t snd_emu10k1_interrupt(int irq, void *dev_id) } outl(orig_status, emu->port + IPR); /* ack all */ } - if (timeout == 1000) - dev_info(emu->card->dev, "emu10k1 irq routine failure\n");
return IRQ_RETVAL(handled); }
We'd try to iterate the voices twice without resetting the pointer. This went unnoticed, because the code isn't actually in use.
Amends commit 27ae958cf6 ("emu10k1 driver - add multichannel device hw:x,3 [2-8/8]").
Signed-off-by: Oswald Buddenhagen oswald.buddenhagen@gmx.de --- sound/pci/emu10k1/irq.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)
diff --git a/sound/pci/emu10k1/irq.c b/sound/pci/emu10k1/irq.c index 7dc803aaa850..a813ef8c2f8d 100644 --- a/sound/pci/emu10k1/irq.c +++ b/sound/pci/emu10k1/irq.c @@ -47,12 +47,13 @@ irqreturn_t snd_emu10k1_interrupt(int irq, void *dev_id) status &= ~(IPR_VOLINCR|IPR_VOLDECR|IPR_MUTE); } if (status & IPR_CHANNELLOOP) { + struct snd_emu10k1_voice *pvoice; int voice; int voice_max = status & IPR_CHANNELNUMBERMASK; u32 val; - struct snd_emu10k1_voice *pvoice = emu->voices;
val = snd_emu10k1_ptr_read(emu, CLIPL, 0); + pvoice = emu->voices; for (voice = 0; voice <= voice_max; voice++) { if (voice == 0x20) val = snd_emu10k1_ptr_read(emu, CLIPH, 0); @@ -68,6 +69,7 @@ irqreturn_t snd_emu10k1_interrupt(int irq, void *dev_id) pvoice++; } val = snd_emu10k1_ptr_read(emu, HLIPL, 0); + pvoice = emu->voices; for (voice = 0; voice <= voice_max; voice++) { if (voice == 0x20) val = snd_emu10k1_ptr_read(emu, HLIPH, 0);
On Thu, 18 May 2023 11:30:44 +0200, Oswald Buddenhagen wrote:
IPR_CHANNELNUMBERMASK cannot be non-zero when IPR_CHANNELLOOP is unset, so join marking them as handled. This logically reverts part of commit f453e20d8a0 ("ALSA update 0.9.3a"), which made the inverse change with no explanation.
Signed-off-by: Oswald Buddenhagen oswald.buddenhagen@gmx.de
Applied all four patches now. Thanks.
Takashi
participants (2)
-
Oswald Buddenhagen -
Takashi Iwai