Received-SPF: pass (google.com: domain of linux-nfs-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: 7bit
MIME-Version: 1.0
From:   "NeilBrown" <neilb@suse.de>
To:     "Chuck Lever" <cel@kernel.org>
Cc:     linux-nfs@vger.kernel.org, "Chuck Lever" <chuck.lever@oracle.com>,
        lorenzo@kernel.org, jlayton@redhat.com, david@fromorbit.com
Subject: Re: [PATCH RFC 0/8] RPC service thread scheduler optimizations
In-reply-to: <168809295522.9283.8453285193952262503@noble.neil.brown.name>
References: <168806401782.1034990.9686296943273298604.stgit@morisot.1015granger.net>,
 <168809295522.9283.8453285193952262503@noble.neil.brown.name>
Date:   Fri, 30 Jun 2023 13:04:44 +1000
Message-id: <168809428416.9283.10675552657236895730@noble.neil.brown.name>
Precedence: bulk

On Fri, 30 Jun 2023, NeilBrown wrote:
> 
>  I agree that an a priori cap on number of threads is not ideal.
>  Have you considered using the xarray to only store busy threads?
>  I think its lookup mechanism mostly relies on a bitmap of present
>  entries, but I'm not completely sure.

It might be event better to use xa_set_mark() and xa_clear_mark() to
manage the busy state.
These are not atomic so you would need an atomic operation in the rqst.

#define XA_MARK_IDLE XA_MARK_1

do {
 rqstp = xa_find(xa, ..., XA_MARK_IDLE);
 if (test_and_set_bit(RQ_BUSY, &rqstp->rq_flags)) {
     xa_clear_mark(xa, rqstp->xa_index, XA_MARK_IDLE);
     break;
 }
} while {rqstp);

xa_find() should be nearly as fast at find_next_bit()
  
NeilBrown