Hoping for a little insight

Hoping for a little insight

[Justin Wojdacki]

Hopefully someone can lend a little insight into what I'm seeing:

I am trying to copy a largish file (well, relative to the flash size)
and after a variable point in the copy, "cp" goes to sleep and never
wakes up again.

My configuration:

Linux Kernel 2.4.16
MTD code as ships w/ above kernel
MIPS Architecture
ST 29W320DB Flash (4 MB)
JFFS2 File System
625000 Byte file (copy of compressed kernel)

I've noticed that a number of SIGHUP's are delivered during the copy,
but they stop about the same time the process is left permanently
asleep. 

Any suggestions/insight/etc?

-- 
-------------------------------------------------
Justin Wojdacki        
justin.wojdacki@analog.com         (408) 350-5032
Communications Processors Group -- Analog Devices

Re: Hoping for a little insight

[David Woodhouse]

justin.wojdacki@analog.com said:
>  I am trying to copy a largish file (well, relative to the flash size)
> and after a variable point in the copy, "cp" goes to sleep and never
> wakes up again. 

SysRq-T.

--
dwmw2

Re: Hoping for a little insight

[Justin Wojdacki]

David Woodhouse wrote:
> 
> justin.wojdacki@analog.com said:
> >  I am trying to copy a largish file (well, relative to the flash size)
> > and after a variable point in the copy, "cp" goes to sleep and never
> > wakes up again.
> 
> SysRq-T.
> 
> --
> dwmw2

Sorry, no luck getting SysRq-T to work. I think the host OS
(WindowsXP) or terminal program (TeraTerm) might be swallowing it. 

If I cat /proc/X/status, I see the following:

Name:   cp
State:  S (sleeping)
Tgid:   30
Pid:    30
PPid:   27
TracerPid:      0
Uid:    0       0       0       0
Gid:    0       0       0       0
FDSize: 32
Groups:
VmSize:      392 kB
VmLck:         0 kB
VmRSS:       104 kB
VmData:       20 kB
VmStk:         4 kB
VmExe:         0 kB
VmLib:         0 kB
SigPnd: 00000000000000000000000000000000
SigBlk: 00000000000000000000000000000000
SigIgn: 00000000000000000000000000000006
SigCgt: 00000000000000000000000000000000
CapInh: 0000000000000000
CapPrm: 00000000fffffeff
CapEff: 00000000fffffeff

-- 
-------------------------------------------------
Justin Wojdacki        
justin.wojdacki@analog.com         (408) 350-5032
Communications Processors Group -- Analog Devices

Re: Hoping for a little insight

[Russ Dill]

On Tue, 2002-10-15 at 16:21, Justin Wojdacki wrote:
> David Woodhouse wrote:
> > 
> > justin.wojdacki@analog.com said:
> > >  I am trying to copy a largish file (well, relative to the flash size)
> > > and after a variable point in the copy, "cp" goes to sleep and never
> > > wakes up again.
> > 
> > SysRq-T.
> > 
> > --
> > dwmw2
> 
> Sorry, no luck getting SysRq-T to work. I think the host OS
> (WindowsXP) or terminal program (TeraTerm) might be swallowing it. 

send a serial break, followed by a t

Re: Hoping for a little insight

[Justin Wojdacki]

Russ Dill wrote:
> 
> >
> > Sorry, no luck getting SysRq-T to work. I think the host OS
> > (WindowsXP) or terminal program (TeraTerm) might be swallowing it.
> 
> send a serial break, followed by a t
> 

Nothing. :(

-- 
-------------------------------------------------
Justin Wojdacki        
justin.wojdacki@analog.com         (408) 350-5032
Communications Processors Group -- Analog Devices

Re: Hoping for a little insight

[Justin Wojdacki]

Is this helpful?

jffs2_reserve_space_gc of 196 bytes for garbage_collect_dnode failed:
-4

-- 
-------------------------------------------------
Justin Wojdacki        
justin.wojdacki@analog.com         (408) 350-5032
Communications Processors Group -- Analog Devices

Re: Hoping for a little insight

[Justin Wojdacki]

FYI for those who might be able to help:

The filesystem has ~2MB free. I'm copying a 600K file from one
directory to another. The filesystem is reporting that it's running
out of space to garbage collect and the write process is hanging. 

-- 
-------------------------------------------------
Justin Wojdacki        
justin.wojdacki@analog.com         (408) 350-5032
Communications Processors Group -- Analog Devices

Re: Hoping for a little insight

[Justin Wojdacki]

Okay, so it looks like far more data is being written to flash than
should be. As an example, a 8205 byte file is occupying 15356 bytes of
actual flash (based on tracking writes). It looks like data is getting
repeatedly written to flash. So, if I have the pattern "ABCD", I see
written to flash:

A
AB
ABC
ABCD

Additionally, I see the flash being written in blocks of 512 bytes. I
have CONFIG_MTD_CFI and CONFIG_MTD_JEDEC_PROBE set in my
configuration, and the following entry added to jedec_probe.c:

	{
                mfr_id: MANUFACTURER_ST,
                dev_id: M29W320DB,   
                name: "ST M29W320DB",
                DevSize: SIZE_4MiB,
                NumEraseRegions: 4,
                regions: { ERASEINFO(16*1024,1),
                           ERASEINFO(8*1024,2),
                           ERASEINFO(32*1024,1),
                           ERASEINFO(64*1024,63)
                }
	}

Is this sufficient? Or should I have added an entry to amd_flash.c
instead and set CONFIG_MTD_AMDSTD? (Some devices are declared in both
jedec_probe.c and amd_flash.c so I'm a little unsure about how this
breaks down). 

-- 
-------------------------------------------------
Justin Wojdacki        
justin.wojdacki@analog.com         (408) 350-5032
Communications Processors Group -- Analog Devices