pipe read returning EAGAIN

[Manuel Bouyer <bouyer%antioche.eu.org@localhost>]

Hello,
I just upgraded a i386 server from netbsd-5 to netbsd-7, and notice that
some processes are using a lot of CPU time.
ktrace shows that such process are in a tight read(2) loop.
Below is an example with nagios but I've seen it with arpwatch too.
netadmin:/tmp#ktrace -p 19040
netadmin:/tmp#fstat -p 19040
USER     CMD          PID   FD MOUNT       INUM MODE         SZ|DV R/W
nagios   nagios     19040   wd /tmp           2 drwxrwxrwt    1024 r
nagios   nagios     19040    0 /          63064 crw-rw-rw-    null r
nagios   nagios     19040    1 /          63064 crw-rw-rw-    null w
nagios   nagios     19040    2 /          63064 crw-rw-rw-    null w
nagios   nagios     19040    3* unix stream  <-> /var/spool/nagios/rw/nagios.qh [using]
nagios   nagios     19040    4* pipe 0xc8298190 <- 0xc8f08c18 rn
nagios   nagios     19040    6* pipe 0xc8f1d91c <- 0xc8299310 rn
netadmin:/tmp#ktrace -C
netadmin:/tmp#kdump |tail
 19040      1 nagios   CALL  read(4,0xbf7fd4e0,0x1000)
 19040      1 nagios   RET   read -1 errno 35 Resource temporarily unavailable
 19040      1 nagios   CALL  read(4,0xbf7fd4e0,0x1000)
 19040      1 nagios   RET   read -1 errno 35 Resource temporarily unavailable
 19040      1 nagios   CALL  read(4,0xbf7fd4e0,0x1000)
 19040      1 nagios   RET   read -1 errno 35 Resource temporarily unavailable
 19040      1 nagios   CALL  read(4,0xbf7fd4e0,0x1000)
 19040      1 nagios   RET   read -1 errno 35 Resource temporarily unavailable
 19040      1 nagios   CALL  read(4,0xbf7fd4e0,0x1000)
 19040      1 nagios   RET   read -1 errno 35 Resource temporarily unavailable


It looks like the read(2) syscall returns a EAGAIN when the caller
expect it to block if there's no data available.

Has anyone else seen this, or has an idea where to look ?

-- 
Manuel Bouyer <bouyer%antioche.eu.org@localhost>
     NetBSD: 26 ans d'experience feront toujours la difference
--