Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
block: Switch from BFQ-v7r7 for 3.10 to BFQ-v7r8 for 3.10
v7r8: . BUGFIX: Let weight-related fields of a bfq_entity be correctly initialized (also) when the I/O priority of the entity is changed before the first request is inserted into the bfq_queue associated to the entity. . BUGFIX: When merging requests belonging to different bfq_queues, avoid repositioning the surviving request. In fact, in this case the repositioning may result in the surviving request being moved across bfq_queues, which would ultimately cause bfq_queues' data structures to become inconsistent. . BUGFIX: When merging requests belonging to the same bfq_queue, reposition the surviving request so that it gets in the correct position, namely the position of the dropped request, instead of always being moved to the head of the FIFO of the bfq_queue (which means to let the request be considered the eldest one). . BUGFIX: Reduce the idling slice for seeky queues only if the scenario is symmetric. This guarantees that also processes associated to seeky queues do receive their reserved share of the throughput. Contributed by Riccardo Pizzetti and Samuele Zecchini. . IMPROVEMENT: Always perform device idling if the scenario is asymmetric in terms of throughput distribution among processes. This extends throughput-distribution guarantees to any process, regardless of the properties of its request pattern and of the request patterns of the other processes, and regardless of whether the device is NCQ-capable. . IMPROVEMENT: Remove the current limitation on the maximum number of in-flight requests allowed for a sync queue (limitation set in place for fairness issues in CFQ, inherited by the first version of BFQ, but made unnecessary by the latest accurate fairness strategies added to BFQ). Removing this limitation enables devices with long internal queues to fill their queues as much as they deem appropriate, also with sync requests. This avoids throughput losses on these devices, because, to achieve a high throughput, they often need to have a high number of requests queued internally. . CODE IMPROVEMENT: Simplify I/O priority change logic by turning it into a single-step procedure instead of a two-step one; improve readability by rethinking the names of the functions involved in changing the I/O priority of a bfq_queue. Signed-off-by: Javier Sayago <admin@lonasdigital.com>
- Loading branch information
Showing
3 changed files
with
77 additions
and
92 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Oops, something went wrong.