[Issue 1000770] New: eCosCentric NAND library

[Issue 1000770] New: eCosCentric NAND library

[bugzilla-daemon]

http://bugzilla.ecoscentric.com/show_bug.cgi?id=1000770

           Summary: eCosCentric NAND library
           Product: eCos
           Version: unknown
          Platform: All
        OS/Version: All
            Status: NEW
          Severity: enhancement
          Priority: normal
         Component: Patches and contributions
        AssignedTo: wry@ecoscentric.com
        ReportedBy: wry@ecoscentric.com
         QAContact: ecos-patches@ecos.sourceware.org
             Class: Advice Request


Created an attachment (id=723)
 --> (http://bugzilla.ecoscentric.com/attachment.cgi?id=723)
Documentation snapshot

As discussed on ecos-maintainers, here is an interim drop of the documentation
for the NAND library we have developed in-house. I intend to have the code
cleaned up and ready to share next week, and will post it to this ticket.


-- 
Configure issuemail: http://bugzilla.ecoscentric.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugzilla.ecoscentric.com/show_bug.cgi?id=1000770


Ross Younger <wry@ecoscentric.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #751 is|0                           |1
           obsolete|                            |




--- Comment #7 from Ross Younger <wry@ecoscentric.com>  2009-06-26 16:16:36 ---
Created an attachment (id=760)
 --> (http://bugzilla.ecoscentric.com/attachment.cgi?id=760)
20090626 snapshot

Significant part-refactor based on feedback received.

NAND: 
* add high-level locking;
* create global pagebuffer;
* be less stack-greedy (use that pagebuffer) so as to not crash in a minimal
config;
* rename nand_app.h to nand.h;
* rename some helper macros for consistency;
* make nand_bbt.h internal;
* move eccdamage test into synth driver;
* add more assertions;

Samsung K9 NAND driver: 
* refactor for NAND changes;
* better define chip-platform interface (created k9fxx08x0x.h); 
* replace calls to diag_printf with NAND_ERROR;

ea2468 HAL:
* refactor for K9 changes;
* add RedBoot heap size requirement;

yaffs: 
* merge to upstream timestamp 2009-05-16;
* omit unneeded YAFFS1 and qsort code;
* replace calls to diag_printf with TOUT;
* further general enhancements and bugfixes;

general: 
* retab files for consistency with eCos sources.


-- 
Configure issuemail: http://bugzilla.ecoscentric.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugzilla.ecoscentric.com/show_bug.cgi?id=1000770


Ross Younger <wry@ecoscentric.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #723 is|0                           |1
           obsolete|                            |
 Attachment #748 is|0                           |1
           obsolete|                            |
 Attachment #760 is|0                           |1
           obsolete|                            |




--- Comment #8 from Ross Younger <wry@ecoscentric.com>  2009-08-26 18:43:37 ---
Created an attachment (id=787)
 --> (http://bugzilla.ecoscentric.com/attachment.cgi?id=787)
Latest snapshot

Here is the latest version of our NAND work. Inside the tarball are a number of
EPKs and patches, for better convenience for those who may wish to use it.

The NAND layer and devices are now presented as a single EPK. YAFFS is a
separate EPK, due to its different license. (I have attached the version number
d20090826 to both of these to reflect the fact that they are date-based
snapshots.)

Board-level support is included as separate patches for the ea2468 and
stm3210e_eval platform HALs. The RedBoot NAND enhancements (unchanged since the
drop of June 26) are in a separate patch.

Summary of changes:

1. Added support for the stm3210e_eval board, incorporating work by Simon
Kallweit, which I have refined into a two-mode low-level driver.

2. Added support for the ST Microelectronics NANDxxxx3A chip (by Simon
Kallweit).

3. Many enhancements to YAFFS:
* Docs expanded with a detailed memory usage calculation (though some recent
work has not yet been documented)
* Added erasenand.c utility
* Update to latest version from Aleph One's CVS repo
* Added small-page device support: added YAFFS1 support (incorporating and
building on Simon Kallweit's work) and YAFFS2 in inbandTags mode; which is used
is configurable.
* Multiple bugfixes: free-space calculation; add tag size sanity check

4. Enhancements to NAND: 
* Fixed corner case bug revealed by extreme stress testing
* Fixed embarrassing BBT initialisation bug
* Added read/write/erase benchmarker "test"

5. NAND synth device: add configurable bad-block injection

6. Samsung K9 chip driver: small optimisations


N.B. Documentation has yet to be written for the above new features, but its
current form can be seen in the eCosPro online docs:

NAND:
http://www.ecoscentric.com/ecospro/doc.cgi/html/ecospro-ref/io-nand-overview.html

NAND drivers:
http://www.ecoscentric.com/ecospro/doc.cgi/html/ecospro-ref/nand-drivers-present.html

YAFFS:
http://www.ecoscentric.com/ecospro/doc.cgi/html/ecospro-ref/fs-yaffs-p.html


-- 
Configure issuemail: http://bugzilla.ecoscentric.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugzilla.ecoscentric.com/show_bug.cgi?id=1000770





--- Comment #9 from Ross Younger <wry@ecoscentric.com>  2009-08-26 19:01:20 ---
(In reply to comment #8)

More recent (part) docs for the synth NAND device are, it turns out, currently
lurking online too:
http://www.ecoscentric.com/ecospro/doc.cgi/html/ref/devs-nand-synth.html

(This needs to be sorted out and the two amalgamated - doing so is on my todo
list.)


-- 
Configure issuemail: http://bugzilla.ecoscentric.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugzilla.ecoscentric.com/show_bug.cgi?id=1000770





--- Comment #10 from Ross Younger <wry@ecoscentric.com>  2009-10-05 16:36:13 ---
Since I last updated this ticket, the docs have been cleaned up a bit and
(most?) anomalies removed. The individual NAND chip drivers now have their own
chapters in the eCosPro reference manual, adjacent to the section for this NAND
library; board-specific configuration options have been moved to the eCosPro EA
LPC2468 and STM3210E HAL port docs.

The NAND driver and synth target links I mentioned earlier in this ticket have
rotted, in favour of the following: 
*
http://www.ecoscentric.com/ecospro/doc.cgi/html/ecospro-ref/nanddev-samsung-k9.html
*
http://www.ecoscentric.com/ecospro/doc.cgi/html/ecospro-ref/devs-nand-st-xxxx3a.html
*
http://www.ecoscentric.com/ecospro/doc.cgi/html/ecospro-ref/devs-nand-synth-ref.html


-- 
Configure issuemail: http://bugzilla.ecoscentric.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1000770


Ross Younger <wry@ecoscentric.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #747 is|0                           |1
           obsolete|                            |




--- Comment #6 from Ross Younger <wry@ecoscentric.com>  2009-06-03 17:30:08 ---
Created an attachment (id=751)
 --> (http://bugs.ecos.sourceware.org/attachment.cgi?id=751)
snapshot

- Remove cyg_nand_init and replace with a static constructor
- Removed all direct use of diag_printf from NAND lib and had it use a
(cdl-config'able) trace function
- Moved (most of) NAND out of libextras.a and into libtarget
- Added RedBoot from anoncvs; added nand utility commands to it
- Sundry bugfixes and tidyups (mostly in YAFFS)


-- 
Configure issuemail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1000770





--- Comment #5 from Ross Younger <wry@ecoscentric.com>  2009-05-28 14:46:50 ---
Created an attachment (id=748)
 --> (http://bugs.ecos.sourceware.org/attachment.cgi?id=748)
YAFFS package docs (HTML)


-- 
Configure issuemail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1000770


Ross Younger <wry@ecoscentric.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #726 is|0                           |1
           obsolete|                            |




--- Comment #4 from Ross Younger <wry@ecoscentric.com>  2009-05-27 18:32:34 ---
Created an attachment (id=747)
 --> (http://bugs.ecos.sourceware.org/attachment.cgi?id=747)
Today's snapshot comprising NAND and YAFFS for anoncvs


-- 
Configure issuemail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1000770





--- Comment #3 from Ross Younger <wry@ecoscentric.com>  2009-05-26 12:57:34 ---
In testing YAFFS on real hardware, it has come (painfully) to my attention that
the NAND library uses too much stack at certain times. (In particular, the
stack overflows on the EA LPC2468 board in a minimal config with
template-default stack sizes.) I will address this in a later drop.


-- 
Configure issuemail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1000770





--- Comment #2 from Ross Younger <wry@ecoscentric.com>  2009-05-18 18:05:51 ---
Created an attachment (id=726)
 --> (http://bugs.ecos.sourceware.org/attachment.cgi?id=726)
NAND repository snapshot

This is the NAND side of my working tree as of today - NAND library, devices
and platform HAL support.

Note: eCos and eCosPro have different HAL board ports for the Embedded Artists
LPC2468 board (ea2468 vs. ea_lpc2468_{16,32}). This export only contains the
HAL from anoncvs, suitably patched; mind the ifdefs in ea_lpc2468_nand.c which
allow the same source to work in both HALs.


-- 
Configure issuemail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.

[Issue 1000770] eCosCentric NAND library

[bugzilla-daemon]

http://bugs.ecos.sourceware.org/show_bug.cgi?id=1000770





--- Comment #1 from Ross Younger <wry@ecoscentric.com>  2009-05-15 18:10:54 ---
I should add that, while I have been working with the in-house maintainers,
these docs - and hence the code - has not yet been properly reviewed by them.
In particular I am aware that there are some contentious issues (logging in
particular) which could well merit a rethink and/or refactor.


-- 
Configure issuemail: http://bugs.ecos.sourceware.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the issue.