Robert Milkowski <[EMAIL PROTECTED]> wrote:

> Hello opensolaris-code,
>
>   U3+patches (Generic_125100-10) - I know, I know - but I belive it is
>   also the case on snv.
>
> # lofiadm
> Block Device             File
> /dev/lofi/1              /home/milek/sol-nv-b69-x86-dvd.iso
>   
> # mount -v | grep lofi
> /dev/lofi/1 on /mnt type hsfs 
> read-only/setuid/devices/noglobal/maplcase/rr/traildot/dev=24c0001 on Mon Aug 
>  6 18:06:55 2007
>
>
> # cd /mnt/Solaris_11/Tools/
> # ./setup_install_server /se3310/jumpstart/os/snv-b69-x86
>
> I can see one 'du' process consuming entire CPU.
>
> # dtrace -n fbt:::entry'{self->vt=vtimestamp;}' -n 
> fbt:::return'/self->vt/[EMAIL PROTECTED](vtimestamp-self->vt);self->vt=0;}' 
> -n tick-5s'{printa(@);exit(0);}'
>
> [...]
>   strncmp                                                     9034200
>   hwblkclr                                                    9044100
>   page_next_scan_large                                       22107700
>   kcage_walk_cage                                            35919800
>   generic_idle_cpu                                           62118000
>   page_numtopp_nolock                                        90173700
>   disp_getwork                                              121844000
>   hs_findhash                                              3012048500
>
> #
> # dtrace -n profile-97'[EMAIL PROTECTED],stack()]=count();}'
>
> [...]
>         1
>               hsfs`hs_findhash+0x2c
>               hsfs`hs_makenode+0x54
>               hsfs`process_dirblock+0x3d0
>               hsfs`hs_dirlook+0x224
>               genunix`fop_lookup+0x28
>               genunix`lookuppnvp+0x344
>               genunix`lookuppnat+0x120
>               genunix`lookupnameat+0x5c
>               genunix`chdir+0x24
>               unix`syscall_trap32+0xcc
>                46
...

> Looks like something in hsfs module is not efficient...

If you like to run tests, please start with a halfway recent hsfs, e.g.:

ftp://ftp.berlios.de/pub/schily/kernel/hsfs/

It seems that you have some ideas why hsfs is slow but you did not name them
explicitely. Could you give us more information?

I know that an important reason for a slow hsfs is the number of seeks
on the medium which will now show up in your tests.....

Last week we had already a discussion on that hsfs needs to do more
caching for meta data to avoid seeks on the medium.

Jörg

-- 
 EMail:[EMAIL PROTECTED] (home) Jörg Schilling D-13353 Berlin
       [EMAIL PROTECTED]                (uni)  
       [EMAIL PROTECTED]     (work) Blog: http://schily.blogspot.com/
 URL:  http://cdrecord.berlios.de/old/private/ ftp://ftp.berlios.de/pub/schily
_______________________________________________
opensolaris-code mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/opensolaris-code

Reply via email to