xfs
[Top] [All Lists]

Re: [PATCH v2 00/12] xfs: kill hard-coded number 128 for transaction spa

To: Jeff Liu <jeff.liu@xxxxxxxxxx>
Subject: Re: [PATCH v2 00/12] xfs: kill hard-coded number 128 for transaction space log reservation
From: Mark Tinguely <tinguely@xxxxxxx>
Date: Fri, 18 Jan 2013 15:40:52 -0600
Cc: xfs@xxxxxxxxxxx
Delivered-to: xfs@xxxxxxxxxxx
In-reply-to: <50EEC680.9040903@xxxxxxxxxx>
References: <50EEC680.9040903@xxxxxxxxxx>
User-agent: Mozilla/5.0 (X11; FreeBSD amd64; rv:9.0) Gecko/20120122 Thunderbird/9.0
On 01/10/13 07:47, Jeff Liu wrote:
Hello,

Here is the v2 patch set of killing hard-coded number 128 which is used to 
indicate
the extra log space reservation for almost all of those transactions.

In this round, I also tried to convert some transactions to pre-calculate out 
the
space log reservation from runtime to mount time so that we can make the code 
related
to xfs_trans_reserve() looks a bit neat and reduce a bit performance 
overhead(basically
can be ignored. :)) IMHO, which were includes: super block quota flags changes,
quota off/end of quota off, adjust quota limits, quota allocations, log dummy1,
log super block counters, log super block units/fields, as well as set 
attributes.


Changes of v2 to v1:
- use xfs_calc_buf_res() to calulate out the space log reservation per item.

Old patches:
v1:
http://www.spinics.net/lists/xfs/msg15499.html

Thanks,
-Jeff

_______________________________________________
xfs mailing list
xfs@xxxxxxxxxxx
http://oss.sgi.com/mailman/listinfo/xfs


Hi Jeff,

I did a quick read of the series and it looks good.

Any reason to have separate constants for:

 XFS_SYNC_ICSBCOUNT_LOG_RES(mp)
 XFS_SYNC_ICSBUNIT_LOG_RES(mp)
 XFS_LOG_DUMMY1_LOG_RES(mp)

since they are the same value and are all superblock operations.

--Mark Tinguely.

<Prev in Thread] Current Thread [Next in Thread>