| Message ID | 20180517180548.6335-1-tiwai@suse.de (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 17.05.2018 20:05, Takashi Iwai wrote:
> Bugzilla: http://bugzilla.opensuse.org/show_bug.cgi?id=1093027
FYI: When developing IOMMU workarounds for the snd_emu10k1 driver
I've also hit bugs in AMD IOMMU kernel driver (as far as I can remember
they only shown for <32-bit DMA masks, that is on Live, but not on
Audigy).
The earliest released kernel to have them fixed was probably 4.15.0.
Maciej
On Fri, 18 May 2018 14:12:32 +0200, Maciej S. Szmigiero wrote: > > On 17.05.2018 20:05, Takashi Iwai wrote: > > Bugzilla: http://bugzilla.opensuse.org/show_bug.cgi?id=1093027 > > FYI: When developing IOMMU workarounds for the snd_emu10k1 driver > I've also hit bugs in AMD IOMMU kernel driver (as far as I can remember > they only shown for <32-bit DMA masks, that is on Live, but not on > Audigy). > > The earliest released kernel to have them fixed was probably 4.15.0. That makes sense, as there have been lots of iova rewrites that happened in 4.15 development. thanks, Takashi
diff --git a/sound/pci/emu10k1/memory.c b/sound/pci/emu10k1/memory.c index 5865f3b90b34..dbc7d8d0e1c4 100644 --- a/sound/pci/emu10k1/memory.c +++ b/sound/pci/emu10k1/memory.c @@ -248,13 +248,13 @@ search_empty(struct snd_emu10k1 *emu, int size) static int is_valid_page(struct snd_emu10k1 *emu, dma_addr_t addr) { if (addr & ~emu->dma_mask) { - dev_err(emu->card->dev, + dev_err_ratelimited(emu->card->dev, "max memory size is 0x%lx (addr = 0x%lx)!!\n", emu->dma_mask, (unsigned long)addr); return 0; } if (addr & (EMUPAGESIZE-1)) { - dev_err(emu->card->dev, "page is not aligned\n"); + dev_err_ratelimited(emu->card->dev, "page is not aligned\n"); return 0; } return 1; @@ -345,7 +345,7 @@ snd_emu10k1_alloc_pages(struct snd_emu10k1 *emu, struct snd_pcm_substream *subst else addr = snd_pcm_sgbuf_get_addr(substream, ofs); if (! is_valid_page(emu, addr)) { - dev_err(emu->card->dev, + dev_err_ratelimited(emu->card->dev, "emu: failure page = %d\n", idx); mutex_unlock(&hdr->block_mutex); return NULL;
The error messages at sanity checks of memory pages tend to repeat too many times once when it hits, and without the rate limit, it may flood and become unreadable. Replace such messages with the *_ratelimited() variant. Bugzilla: http://bugzilla.opensuse.org/show_bug.cgi?id=1093027 Signed-off-by: Takashi Iwai <tiwai@suse.de> --- sound/pci/emu10k1/memory.c | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-)