Document cached_read_{data,meta,journal} parameters

docs/config/osd.en.md

@@ -31,6 +31,9 @@ them, even without restarting by updating configuration in etcd.
 - [max_flusher_count](#max_flusher_count)
 - [inmemory_metadata](#inmemory_metadata)
 - [inmemory_journal](#inmemory_journal)
+- [cached_read_data](#cached_read_data)
+- [cached_read_meta](#cached_read_meta)
+- [cached_read_journal](#cached_read_journal)
 - [journal_sector_buffer_count](#journal_sector_buffer_count)
 - [journal_no_same_sector_overwrites](#journal_no_same_sector_overwrites)
 - [throttle_small_writes](#throttle_small_writes)
@@ -255,6 +258,46 @@ is typically very small because it's sufficient to have 16-32 MB journal
 for SSD OSDs. However, in theory it's possible that you'll want to turn it
 off for hybrid (HDD+SSD) OSDs with large journals on quick devices.
 
+## cached_read_data
+
+- Type: boolean
+- Default: false
+
+Read data through Linux page cache, i.e. use a file descriptor opened without
+O_DIRECT for data reads. May improve read performance for frequently accessed
+data if it fits in RAM. Memory in page cache is shared by all processes and
+not accounted in OSD memory consumption.
+
+## cached_read_meta
+
+- Type: boolean
+- Default: false
+
+Read metadata through Linux page cache. May be beneficial when checksums
+are enabled and [inmemory_metadata](#inmemory_metadata) is disabled, because
+in this case metadata blocks are read from disk to verify checksums on every
+read request and caching them may reduce this extra read load.
+
+Absolutely pointless to enable with enabled inmemory_metadata because all
+metadata is kept in memory anyway, and likely pointless without checksums,
+because in that case, metadata blocks are read from disk only during journal
+flushing.
+
+If the same device is used for data and metadata, enabling [cached_read_data](#cached_read_data)
+also enables this parameter, given that it isn't turned off explicitly.
+
+## cached_read_journal
+
+- Type: boolean
+- Default: false
+
+Read buffered data from journal through Linux page cache. Does not have sense
+without disabling [inmemory_journal](#inmemory_journal), which, again, is
+enabled by default.
+
+If the same device is used for metadata and journal, enabling [cached_read_meta](#cached_read_meta)
+also enables this parameter, given that it isn't turned off explicitly.
+
 ## journal_sector_buffer_count
 
 - Type: integer