github/ocf
2
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Refactor cache mngt stop/detach for better error handling

Browse Source 
_ocf_mngt_cache_unplug context is now provided by the caller.
This way _ocf_mngt_cache_unplug returns only non-critical (cache write)
errors, allowing stop/detach operation to always proceed and optionally
finish with error. This eliminates the need for rolling back previous
stop/detach operations, which might turn out to be impossible e.g.
under memory pressure.

Signed-off-by: Adam Rutkowski <adam.j.rutkowski@intel.com>
This commit is contained in:
Adam Rutkowski 2019-04-23 15:51:50 -04:00
parent bf879c4b09
commit 26579e53cd
1 changed files with 42 additions and 36 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

78
src/mngt/ocf_mngt_cache.c
View File

@ -1845,7 +1845,6 @@ static void _ocf_mngt_cache_unplug_complete(void *priv, int error)
_ocf_mngt_init_attached_nonpersistent(cache);
context->cmpl(context->priv, error ? -OCF_ERR_WRITE_CACHE : 0);
env_vfree(context);
}
/**
@ -1858,22 +1857,16 @@ static void _ocf_mngt_cache_unplug_complete(void *priv, int error)
* clean shutdown in metadata and flush all containers.
* - false if the device is to be detached from cache - loading
* metadata from this device will not be possible.
* @param context - context for this call, must be zeroed
* @param cmpl Completion callback
* @param priv Completion context
*/
static void _ocf_mngt_cache_unplug(ocf_cache_t cache, bool stop,
struct _ocf_mngt_cache_unplug_context *context,
_ocf_mngt_cache_unplug_end_t cmpl, void *priv)
{
struct _ocf_mngt_cache_unplug_context *context;
ENV_BUG_ON(stop && cache->conf_meta->core_count != 0);
context = env_vzalloc(sizeof(*context));
if (!context) {
cmpl(priv, -OCF_ERR_NO_MEM);
return;
}
context->cmpl = cmpl;
context->priv = priv;
context->cache = cache;
@ -1969,6 +1962,12 @@ void ocf_mngt_cache_load(ocf_cache_t cache,
}
struct ocf_mngt_cache_stop_context {
/* unplug context - this is private structure of _ocf_mngt_cache_unplug,
* it is member of stop context only to reserve memory in advance for
* _ocf_mngt_cache_unplug, eliminating the possibility of ENOMEM error
* at the point where we are effectively unable to handle it */
struct _ocf_mngt_cache_unplug_context unplug_context;
ocf_mngt_cache_stop_end_t cmpl;
void *priv;
ocf_pipeline_t pipeline;
@ -2017,15 +2016,10 @@ static void ocf_mngt_cache_stop_unplug_complete(void *priv, int error)
{
struct ocf_mngt_cache_stop_context *context = priv;
/* short-circut execution in case of critical error */
if (error && error != -OCF_ERR_WRITE_CACHE) {
ocf_pipeline_finish(context->pipeline, error);
return;
}
/* in case of non-critical (disk write) error just remember its value */
if (error)
if (error) {
ENV_BUG_ON(error != -OCF_ERR_WRITE_CACHE);
context->cache_write_error = error;
}
ocf_pipeline_next(context->pipeline);
}
@ -2041,7 +2035,7 @@ static void ocf_mngt_cache_stop_unplug(ocf_pipeline_t pipeline,
return;
}
_ocf_mngt_cache_unplug(cache, true,
_ocf_mngt_cache_unplug(cache, true, &context->unplug_context,
ocf_mngt_cache_stop_unplug_complete, context);
}
@ -2077,14 +2071,17 @@ static void ocf_mngt_cache_stop_finish(ocf_pipeline_t pipeline,
env_bit_set(ocf_cache_state_running, &cache->cache_state);
}
if (context->cache_write_error) {
ocf_log(ctx, log_warn, "Stopped cache %s with errors\n",
context->cache_name);
} else if (error) {
ocf_log(ctx, log_err, "Stopping cache %s failed\n",
context->cache_name);
if (!error) {
if (!context->cache_write_error) {
ocf_log(ctx, log_info,
"Cache %s successfully stopped\n",
context->cache_name);
} else {
ocf_log(ctx, log_warn, "Stopped cache %s with errors\n",
context->cache_name);
}
} else {
ocf_log(ctx, log_info, "Cache %s successfully stopped\n",
ocf_log(ctx, log_err, "Stopping cache %s failed\n",
context->cache_name);
}
@ -2334,10 +2331,17 @@ int ocf_mngt_cache_get_fallback_pt_error_threshold(ocf_cache_t cache,
}
struct ocf_mngt_cache_detach_context {
/* unplug context - this is private structure of _ocf_mngt_cache_unplug,
* it is member of detach context only to reserve memory in advance for
* _ocf_mngt_cache_unplug, eliminating the possibility of ENOMEM error
* at the point where we are effectively unable to handle it */
struct _ocf_mngt_cache_unplug_context unplug_context;
ocf_mngt_cache_detach_end_t cmpl;
void *priv;
ocf_pipeline_t pipeline;
ocf_cache_t cache;
int cache_write_error;
};
static void ocf_mngt_cache_detach_flush_cmpl(ocf_cache_t cache,
@ -2404,8 +2408,8 @@ static void ocf_mngt_cache_detach_unplug_complete(void *priv, int error)
struct ocf_mngt_cache_detach_context *context = priv;
if (error) {
ocf_pipeline_finish(context->pipeline, error);
return;
ENV_BUG_ON(error != -OCF_ERR_WRITE_CACHE);
context->cache_write_error = error;
}
ocf_pipeline_next(context->pipeline);
@ -2419,7 +2423,7 @@ static void ocf_mngt_cache_detach_unplug(ocf_pipeline_t pipeline,
/* Do the actual detach - deinit cacheline metadata,
* stop cleaner thread and close cache bottom device */
_ocf_mngt_cache_unplug(cache, false,
_ocf_mngt_cache_unplug(cache, false, &context->unplug_context,
ocf_mngt_cache_detach_unplug_complete, context);
}
@ -2432,18 +2436,20 @@ static void ocf_mngt_cache_detach_finish(ocf_pipeline_t pipeline,
ocf_refcnt_unfreeze(&cache->dirty);
if (!error) {
ocf_cache_log(cache, log_info, "Successfully detached\n");
} else {
if (error == -OCF_ERR_WRITE_CACHE) {
ocf_cache_log(cache, log_warn,
"Detached cache with errors\n");
if (!context->cache_write_error) {
ocf_cache_log(cache, log_info,
"Device successfully detached\n");
} else {
ocf_cache_log(cache, log_err,
"Detaching cache failed\n");
ocf_cache_log(cache, log_warn,
"Device detached with errors\n");
}
} else {
ocf_cache_log(cache, log_err,
"Detaching device failed\n");
}
context->cmpl(cache, context->priv, error);
context->cmpl(cache, context->priv,
error ?: context->cache_write_error);
ocf_pipeline_destroy(context->pipeline);
}