Bug#611314: SunFire V120 and sym53c8xx

2011-02-20 Thread Richard Mortimer



On 19/02/2011 23:20, Jurij Smakov wrote:

On Sat, Feb 19, 2011 at 11:06:04PM +, Jurij Smakov wrote:

On Thu, Feb 10, 2011 at 10:49:39AM +, Richard Mortimer wrote:

[Cc'd to 611...@bugs.debian.org]


...

I tested with your netboot image and it exhibits the same behaviour
as the official installer. The sym53c8xx driver does get loaded
automatically. I'm pretty sure that this is a timing issue. I
confirmed that all of the /dev/sd* files were present along with
/dev/disk/* entries.


Ok, I think the magic happens in the disk_found() function in
disk-detect.sh [0]. It looks like it will retry up to 3 times to see
whether devices have showed up (with list-devices disk), sleeping 2
seconds between attempts... If it takes up to 7 seconds for driver to
initialize itself, we might be cutting it just a bit too short.
Unfortunately, I don't see how we can test it easily, as neither
netboot image nor miniiso contain the udebs, they are downloaded off
the network.


Hm, if expert install mode offers an opportunity to drop to shell
after udebs are downloaded, then we can probably replace the udeb
containing disk-detect.sh with a modified one. Can you check whether
it's possible?


Yeah I can use Go Back at Set up users and passwords to get to the 
installer main menu and then drop to a shell.


I can likely modify any scripts in place if you give me pointers to what 
you would like to change/test.


To get things started I tried changing the number of retries from 3 to 8 
with


sed -i 's/1 2 3/1 2 3 4 5 6 7 8/' /bin/disk-detect

This detected the disks just fine.

Regards

Richard



--
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#611314: SunFire V120 and sym53c8xx

2011-02-20 Thread Jurij Smakov
On Sun, Feb 20, 2011 at 02:28:42PM +, Richard Mortimer wrote:
 
 
 On 19/02/2011 23:20, Jurij Smakov wrote:
 On Sat, Feb 19, 2011 at 11:06:04PM +, Jurij Smakov wrote:
 On Thu, Feb 10, 2011 at 10:49:39AM +, Richard Mortimer wrote:
 [Cc'd to 611...@bugs.debian.org]
 
 ...
 I tested with your netboot image and it exhibits the same behaviour
 as the official installer. The sym53c8xx driver does get loaded
 automatically. I'm pretty sure that this is a timing issue. I
 confirmed that all of the /dev/sd* files were present along with
 /dev/disk/* entries.
 
 Ok, I think the magic happens in the disk_found() function in
 disk-detect.sh [0]. It looks like it will retry up to 3 times to see
 whether devices have showed up (with list-devices disk), sleeping 2
 seconds between attempts... If it takes up to 7 seconds for driver to
 initialize itself, we might be cutting it just a bit too short.
 Unfortunately, I don't see how we can test it easily, as neither
 netboot image nor miniiso contain the udebs, they are downloaded off
 the network.
 
 Hm, if expert install mode offers an opportunity to drop to shell
 after udebs are downloaded, then we can probably replace the udeb
 containing disk-detect.sh with a modified one. Can you check whether
 it's possible?
 
 Yeah I can use Go Back at Set up users and passwords to get to the
 installer main menu and then drop to a shell.
 
 I can likely modify any scripts in place if you give me pointers to
 what you would like to change/test.
 
 To get things started I tried changing the number of retries from 3
 to 8 with
 
 sed -i 's/1 2 3/1 2 3 4 5 6 7 8/' /bin/disk-detect

Thanks, that's the confirmation I was looking for. We are going to 
push a patch for this to 6.0.1 as it is fairly non-controversial.
 
 This detected the disks just fine.
 
 Regards
 
 Richard

-- 
Jurij Smakov   ju...@wooyd.org
Key: http://www.wooyd.org/pgpkey/  KeyID: C99E03CC



-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#611314: SunFire V120 and sym53c8xx (was Re: Unofficial Squeeze installer image available)

2011-02-19 Thread Jurij Smakov
On Thu, Feb 10, 2011 at 10:49:39AM +, Richard Mortimer wrote:
 [Cc'd to 611...@bugs.debian.org]
 
 On 10/02/2011 08:36, Jurij Smakov wrote:
 On Wed, Feb 09, 2011 at 11:07:49PM +, Richard Mortimer wrote:
 On 09/02/2011 21:43, Jurij Smakov wrote:
 ... snip ...
 
 http://www.wooyd.org/debian/squeeze/
 
 ...
 
 Sure, I have just uploaded a netboot image to the same location.
 
 If you have encountered problems during Squeeze installation, please
 test this image and report the results, as a positive confirmation of
 fixes on a variety of different systems is essential for getting all
 these fixes into the first Squeeze point release.
 
 I did spot a couple of issues with the installer when I gave squeeze
 a try a couple of weeks back. I only got chance to file an
 installation report about one of them
 http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=611314
 Its not a massive show stopper and has an easy workaround but if
 that can be fixed then it will make others lives easier.
 
 Yeah, this bug was pointed out to me just yesterday. I don't know
 what's going on there yet, it appears that the PCI IDs of your
 card match the ones which are supported by the driver, so if module is
 not loaded at all there is probably something wrong with udev. At a
 point where it says it cannot detect the disks, can you drop back to
 shell and see whether the driver actually isn't loaded (as opposed to
 being loaded but failing to find the disks anyway)? Maybe there is
 something relevant showing up in dmesg or /var/log/messages?
 
 If not, I'll try to come up with some udev debugging commands to try
 and understand why this detection is failing.
 
 I tested with your netboot image and it exhibits the same behaviour
 as the official installer. The sym53c8xx driver does get loaded
 automatically. I'm pretty sure that this is a timing issue. I
 confirmed that all of the /dev/sd* files were present along with
 /dev/disk/* entries.

Ok, I think the magic happens in the disk_found() function in 
disk-detect.sh [0]. It looks like it will retry up to 3 times to see 
whether devices have showed up (with list-devices disk), sleeping 2 
seconds between attempts... If it takes up to 7 seconds for driver to 
initialize itself, we might be cutting it just a bit too short. 
Unfortunately, I don't see how we can test it easily, as neither 
netboot image nor miniiso contain the udebs, they are downloaded off 
the network.

[0] 
http://git.debian.org/?p=d-i/hw-detect.git;a=blob_plain;f=disk-detect.sh;hb=HEAD

 Looking at the dmesg output I can see a 7 second delay whilst the
 driver/scsi subsystem settles itself down. (I also see a similar
 issue when I boot after install with raid/lvm setup - but both
 rootdelay  the scsi probe sync option fixes that). After dropping
 to the installer shell I can re-enter the Detect disks screen and
 it find them without problem.
 
 Regards
 
 Richard
 
 
 Full console output below just in case it is useful.
 
 
 
 BusyBox v1.17.1 (Debian 1:1.17.1-8) built-in shell (ash)
 Enter 'help' for a list of built-in commands.
 
 ~ #
 ~ # lsmod
 Module  Size  Used by
 sd_mod 31548  0
 crc_t10dif  1292  1 sd_mod
 ata_generic 3199  0
 libata147511  1 ata_generic
 sym53c8xx  71569  0
 scsi_transport_spi 20705  1 sym53c8xx
 alim15x34002  0
 usb_storage41983  0
 scsi_mod  132037  5
 sd_mod,libata,sym53c8xx,scsi_transport_spi,usb_storage
 ohci_hcd   18603  0
 uhci_hcd   20343  0
 ehci_hcd   33777  0
 sungem 25663  0
 usbcore   117062  5 usb_storage,ohci_hcd,uhci_hcd,ehci_hcd
 nls_base6649  1 usbcore
 sungem_phy  9430  1 sungem
 ~ #
 ~ # ls /dev
 block   shm tty42
 bsg stderr  tty43
 bus stdin   tty44
 charstdout  tty45
 console tty tty46
 coretty0tty47
 cpu_dma_latency tty1tty48
 disktty10   tty49
 fd  tty11   tty5
 fulltty12   tty50
 input   tty13   tty51
 kmsgtty14   tty52
 log tty15   tty53
 loop0   tty16   tty54
 mdesc   tty17   tty55
 mem tty18   tty56
 net tty19   tty57
 network_latency tty2tty58
 network_throughput  tty20   tty59
 nulltty21   tty6
 openpromtty22   tty60
 porttty23   tty61
 ppp tty24   tty62
 psaux   tty25   tty63
 ptmxtty26   tty7
 pts   

Bug#611314: SunFire V120 and sym53c8xx (was Re: Unofficial Squeeze installer image available)

2011-02-19 Thread Jurij Smakov
On Sat, Feb 19, 2011 at 11:06:04PM +, Jurij Smakov wrote:
 On Thu, Feb 10, 2011 at 10:49:39AM +, Richard Mortimer wrote:
  [Cc'd to 611...@bugs.debian.org]
  
  On 10/02/2011 08:36, Jurij Smakov wrote:
  On Wed, Feb 09, 2011 at 11:07:49PM +, Richard Mortimer wrote:
  On 09/02/2011 21:43, Jurij Smakov wrote:
  ... snip ...
  
  http://www.wooyd.org/debian/squeeze/
  
  ...
  
  Sure, I have just uploaded a netboot image to the same location.
  
  If you have encountered problems during Squeeze installation, please
  test this image and report the results, as a positive confirmation of
  fixes on a variety of different systems is essential for getting all
  these fixes into the first Squeeze point release.
  
  I did spot a couple of issues with the installer when I gave squeeze
  a try a couple of weeks back. I only got chance to file an
  installation report about one of them
  http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=611314
  Its not a massive show stopper and has an easy workaround but if
  that can be fixed then it will make others lives easier.
  
  Yeah, this bug was pointed out to me just yesterday. I don't know
  what's going on there yet, it appears that the PCI IDs of your
  card match the ones which are supported by the driver, so if module is
  not loaded at all there is probably something wrong with udev. At a
  point where it says it cannot detect the disks, can you drop back to
  shell and see whether the driver actually isn't loaded (as opposed to
  being loaded but failing to find the disks anyway)? Maybe there is
  something relevant showing up in dmesg or /var/log/messages?
  
  If not, I'll try to come up with some udev debugging commands to try
  and understand why this detection is failing.
  
  I tested with your netboot image and it exhibits the same behaviour
  as the official installer. The sym53c8xx driver does get loaded
  automatically. I'm pretty sure that this is a timing issue. I
  confirmed that all of the /dev/sd* files were present along with
  /dev/disk/* entries.
 
 Ok, I think the magic happens in the disk_found() function in 
 disk-detect.sh [0]. It looks like it will retry up to 3 times to see 
 whether devices have showed up (with list-devices disk), sleeping 2 
 seconds between attempts... If it takes up to 7 seconds for driver to 
 initialize itself, we might be cutting it just a bit too short. 
 Unfortunately, I don't see how we can test it easily, as neither 
 netboot image nor miniiso contain the udebs, they are downloaded off 
 the network.

Hm, if expert install mode offers an opportunity to drop to shell 
after udebs are downloaded, then we can probably replace the udeb 
containing disk-detect.sh with a modified one. Can you check whether 
it's possible?
-- 
Jurij Smakov   ju...@wooyd.org
Key: http://www.wooyd.org/pgpkey/  KeyID: C99E03CC



-- 
To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org
with a subject of unsubscribe. Trouble? Contact listmas...@lists.debian.org



Bug#611314: SunFire V120 and sym53c8xx (was Re: Unofficial Squeeze installer image available)

2011-02-10 Thread Richard Mortimer

[Cc'd to 611...@bugs.debian.org]

On 10/02/2011 08:36, Jurij Smakov wrote:

On Wed, Feb 09, 2011 at 11:07:49PM +, Richard Mortimer wrote:

On 09/02/2011 21:43, Jurij Smakov wrote:

... snip ...


http://www.wooyd.org/debian/squeeze/


...


Sure, I have just uploaded a netboot image to the same location.


If you have encountered problems during Squeeze installation, please
test this image and report the results, as a positive confirmation of
fixes on a variety of different systems is essential for getting all
these fixes into the first Squeeze point release.


I did spot a couple of issues with the installer when I gave squeeze
a try a couple of weeks back. I only got chance to file an
installation report about one of them
http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=611314
Its not a massive show stopper and has an easy workaround but if
that can be fixed then it will make others lives easier.


Yeah, this bug was pointed out to me just yesterday. I don't know
what's going on there yet, it appears that the PCI IDs of your
card match the ones which are supported by the driver, so if module is
not loaded at all there is probably something wrong with udev. At a
point where it says it cannot detect the disks, can you drop back to
shell and see whether the driver actually isn't loaded (as opposed to
being loaded but failing to find the disks anyway)? Maybe there is
something relevant showing up in dmesg or /var/log/messages?

If not, I'll try to come up with some udev debugging commands to try
and understand why this detection is failing.


I tested with your netboot image and it exhibits the same behaviour as 
the official installer. The sym53c8xx driver does get loaded 
automatically. I'm pretty sure that this is a timing issue. I confirmed 
that all of the /dev/sd* files were present along with /dev/disk/* entries.


Looking at the dmesg output I can see a 7 second delay whilst the 
driver/scsi subsystem settles itself down. (I also see a similar issue 
when I boot after install with raid/lvm setup - but both rootdelay  the 
scsi probe sync option fixes that). After dropping to the installer 
shell I can re-enter the Detect disks screen and it find them without 
problem.


Regards

Richard


Full console output below just in case it is useful.



BusyBox v1.17.1 (Debian 1:1.17.1-8) built-in shell (ash)
Enter 'help' for a list of built-in commands.

~ #
~ # lsmod
Module  Size  Used by
sd_mod 31548  0
crc_t10dif  1292  1 sd_mod
ata_generic 3199  0
libata147511  1 ata_generic
sym53c8xx  71569  0
scsi_transport_spi 20705  1 sym53c8xx
alim15x34002  0
usb_storage41983  0
scsi_mod  132037  5 
sd_mod,libata,sym53c8xx,scsi_transport_spi,usb_storage

ohci_hcd   18603  0
uhci_hcd   20343  0
ehci_hcd   33777  0
sungem 25663  0
usbcore   117062  5 usb_storage,ohci_hcd,uhci_hcd,ehci_hcd
nls_base6649  1 usbcore
sungem_phy  9430  1 sungem
~ #
~ # ls /dev
block   shm tty42
bsg stderr  tty43
bus stdin   tty44
charstdout  tty45
console tty tty46
coretty0tty47
cpu_dma_latency tty1tty48
disktty10   tty49
fd  tty11   tty5
fulltty12   tty50
input   tty13   tty51
kmsgtty14   tty52
log tty15   tty53
loop0   tty16   tty54
mdesc   tty17   tty55
mem tty18   tty56
net tty19   tty57
network_latency tty2tty58
network_throughput  tty20   tty59
nulltty21   tty6
openpromtty22   tty60
porttty23   tty61
ppp tty24   tty62
psaux   tty25   tty63
ptmxtty26   tty7
pts tty27   tty8
random  tty28   tty9
rtc0tty29   ttyS0
sda tty3ttyS1
sda1tty30   urandom
sda2tty31   vcs
sda3tty32   vcs1
sda7tty33   vcs2
sdb tty34   vcs3
sdb1tty35   vcs4
sdb2tty36   vcsa
sdb3tty37   vcsa1
sdb4tty38   vcsa2
sdb5tty39   vcsa3
sdb6tty4vcsa4
sdb7tty40