[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-scsi
Subject:    Re: [PATCH V4 2/3] scsi: core: avoid to pre-allocate big chunk for protection meta data
From:       Bart Van Assche <bvanassche () acm ! org>
Date:       2019-04-29 18:15:52
Message-ID: 1556561752.161891.165.camel () acm ! org
[Download RAW message or body]

On Sun, 2019-04-28 at 15:39 +-0800, Ming Lei wrote:
+AD4 Now scsi+AF8-mq+AF8-setup+AF8-tags() pre-allocates a big buffer for protection
+AD4 sg entries, and the buffer size is scsi+AF8-mq+AF8-sgl+AF8-size().
+AD4 
+AD4 This way isn't correct, scsi+AF8-mq+AF8-sgl+AF8-size() is used to pre-allocate
+AD4 sg entries for IO data. And the protection data buffer is much less,
+AD4 for example, one 512byte sector needs 8byte protection data, and
+AD4 the max sector number for one request is 2560(BLK+AF8-DEF+AF8-MAX+AF8-SECTORS),
+AD4 so the max protection data size is just 20k.
+AD4 
+AD4 The usual case is that one bio builds one single bip segment. Attribute
+AD4 to bio split, bio merge is seldom done for big IO, and it is only done
+AD4 in case of small bios. And protection data segment number is usually
+AD4 same with bio count in the request, so the number won't be very big,
+AD4 and allocating from slab is fast enough.
+AD4 
+AD4 Reduce to pre-allocate one sg entry for protection data, and switch
+AD4 to runtime allocation in case that the protection data segment number
+AD4 is bigger than 1. Then we can save huge pre-alocation, for example,
+AD4 500

Reviewed-by: Bart Van Assche +ADw-bvanassche+AEA-acm.org+AD4


[prev in list] [next in list] [prev in thread] [next in thread]