ALSA: pcmtest: Extend error injection, decrease buffer filling overhead
Subsystems (2)
Ivan Orlov
maintainerPATCH 1/1 ALSA: pcmtest: Extend error injection, decrease buffer filling overhead
Extend 'pcmtest' virtual driver with 'open' callback error injection functionality, as it already can inject errors into other PCM callbacks.
Fix the driver to use already defined variables where it is possible.
Additionally, decrease the buffer filling overhead with conditional reminder calculation in the 'inc_buf_pos' inline function.
--- sound/drivers/pcmtest.c | 16 ++++++++++++---- 1 file changed, 12 insertions(+), 4 deletions(-)
sound/drivers/pcmtest.c +12 -4
@@ -65,6 +65,7 @@ static int inject_delay;static bool inject_hwpars_err;static bool inject_prepare_err;static bool inject_trigger_err; +static bool inject_open_err;static short fill_mode =FILL_MODE_PAT ;
@@ -88,6 +89,9 @@ module_param(inject_prepare_err, bool, 0600);MODULE_PARM_DESC (inject_prepare_err,"Inject EINVAL error in the 'prepare' callback" );module_param (inject_trigger_err, bool,0600 );MODULE_PARM_DESC (inject_trigger_err,"Inject EINVAL error in the 'trigger' callback" ); +module_param (inject_open_err, bool,0600 ); +MODULE_PARM_DESC (inject_open_err,"Inject EBUSY error in the 'open' callback" ); +struct pcmtst {struct snd_pcm *pcm;
@@ -140,7 +144,8 @@ static inline void inc_buf_pos(struct pcmtst_buf_iter *v_iter, size_t by, size_t { v_iter->total_bytes += by; v_iter->buf_pos += by; - v_iter->buf_pos %= bytes; +if (v_iter->buf_pos >= bytes) + v_iter->buf_pos %= bytes; }/*
@@ -196,10 +201,10 @@ static void check_buf_block_ni(struct pcmtst_buf_iter *v_iter, struct snd_pcm_ruu8 current_byte;for (i =0 ; i < v_iter->b_rw; i++) { - current_byte = runtime->dma_area[buf_pos_n (v_iter, channels, i % channels)]; + ch_num = i % channels; + current_byte = runtime->dma_area[buf_pos_n (v_iter, channels, ch_num)];if (!current_byte)break ; - ch_num = i % channels;if (current_byte != patt_bufs[ch_num].buf[(v_iter->total_bytes / channels) % patt_bufs[ch_num].len]) { v_iter->is_buf_corrupted = true;
@@ -239,7 +244,7 @@ static void fill_block_pattern_n(struct pcmtst_buf_iter *v_iter, struct snd_pcm_for (i =0 ; i < v_iter->b_rw; i++) { ch_num = i % channels; - runtime->dma_area[buf_pos_n (v_iter, channels, i % channels)] = + runtime->dma_area[buf_pos_n (v_iter, channels, ch_num)] = patt_bufs[ch_num].buf[(v_iter->total_bytes / channels) % patt_bufs[ch_num].len];inc_buf_pos (v_iter,1 , runtime->dma_bytes);
@@ -364,6 +369,9 @@ static int snd_pcmtst_pcm_open(struct snd_pcm_substream *substream)struct snd_pcm_runtime *runtime = substream->runtime;struct pcmtst_buf_iter *v_iter; +if (inject_open_err) +return -EBUSY ; + v_iter =kzalloc (sizeof (*v_iter),GFP_KERNEL );if (!v_iter)return -ENOMEM ;
Signature · 1 line
2.34.1
Quoted 10 lines · Ivan Orlov
Extend 'pcmtest' virtual driver with 'open' callback error injection functionality, as it already can inject errors into other PCM callbacks.
Fix the driver to use already defined variables where it is possible.
Additionally, decrease the buffer filling overhead with conditional reminder calculation in the '
inc_buf_pos' inline function.Signed-off-by Ivan Orlov <[email protected]>
Please avoid mixing different changes in a single patch.
This patch does three completely different things, and they should be split.
- New inject_open_err parameter
- Optimization of inc_buf_pos()
- Optimization of check_buf_block_ni() and fill_block_pattern_n()
The latter two could be put in a single patch as minor optimizations, but the introduction of a new option doesn't fit with the rest.
thanks,
Takashi
Ivan Orlov
maintainerQuoted 22 lines · Takashi Iwai
Quoted 10 lines · Ivan Orlov
Extend 'pcmtest' virtual driver with 'open' callback error injection functionality, as it already can inject errors into other PCM callbacks.
Fix the driver to use already defined variables where it is possible.
Additionally, decrease the buffer filling overhead with conditional reminder calculation in the '
inc_buf_pos' inline function.Signed-off-by Ivan Orlov <[email protected]>Please avoid mixing different changes in a single patch.
This patch does three completely different things, and they should be split.
- New
inject_open_errparameter
- Optimization ofinc_buf_pos()
- Optimization ofcheck_buf_block_ni()andfill_block_pattern_n()
Quoted 7 lines
The latter two could be put in a single patch as minor optimizations, but the introduction of a new option doesn't fit with the rest.
thanks,
Takashi
Hi Takashi,
Thank you for the review. I'll split the patch and send the changes again.
Kind regards,
Ivan Orlov