From: Wolfgang Denk Date: Sun, 17 May 2009 16:52:28 +0200 Subject: [U-Boot] [RFC/PATCH] jffs2/mtdparts: Fix problem with usage from JFFS2 and MTDPARTS together In-Reply-To: <1242468262-25814-1-git-send-email-sr@denx.de> References: <1242468262-25814-1-git-send-email-sr@denx.de> Message-ID: <20090517145228.18DF5832E416@gemini.denx.de> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: u-boot@lists.denx.de Dear Stefan, In message <1242468262-25814-1-git-send-email-sr@denx.de> you wrote: > Currently using JFFS2 with MTDPARTS enabled doesn't work. This is because > mtdparts_init() is available in both files, cmd_mtdparts.c and > cmd_jffs2.c. Please note that in the original cmd_jffs2.c file (before > the jffs2/mtdparts command/file split those 2 different versions > already existed. So this is nothing new. The main problem is that the > variables "current_dev" and "current_partnum" are declared in both > files now. This doesn't work. I wonder how the mtdparts code is supposed to work at all. I was running some more tests on a TQM823L boiard, which has two banks of NOR flash, each 4 MiB, mapped at 0x40000000 and 0x40400000, respective, to form a contiguous area of 8 MiB from 0x40000000...0x407FFFFF. This is also what the default MTDPARTS setting assumes: mtdparts=TQM8xxL-0:256k(u-boot),128k(dtb),1664k(kernel),2m(rootfs),4m(data) However, it doesn't work. With DEBUGF enabled and an additional debug output, I see this: => mtdparts default ---mtdparts_init--- last_ids : env_ids : nor0=TQM8xxL-0 last_parts: env_parts : mtdparts=TQM8xxL-0:256k(u-boot),128k(dtb),1664k(kernel),2m(rootfs),4m(data) last_partition : env_partition : ---parse_mtdids--- mtdids = nor0=TQM8xxL-0 + id nor0 4194304 bytes TQM8xxL-0 ---parse_mtdparts--- mtdparts = mtdparts=TQM8xxL-0:256k(u-boot),128k(dtb),1664k(kernel),2m(rootfs),4m(data) --- current_save --- => partition NULL --- index partitions --- => mtddevnum NULL => mtddevname NULL ===device_parse=== --- id_find_by_mtd_id: 'TQM8xxL-0' (len = 9) entry: 'TQM8xxL-0' (len = 9) dev type = 1 (nor), dev num = 0, mtd-id = TQM8xxL-0 parsing partitions 256k(u-boot),128k(dtb),1664k(kernel),2m(rootfs),4m(data) + partition: name u-boot size 0x00040000 offset 0xffffffff mask flags 0 part_validate: id: TQM8xxL-0 p->off: 00000000 p->sz: 00040000 end: 40000 id->sz: 00400000 + partition: name dtb size 0x00020000 offset 0xffffffff mask flags 0 part_validate: id: TQM8xxL-0 p->off: 00040000 p->sz: 00020000 end: 60000 id->sz: 00400000 + partition: name kernel size 0x001a0000 offset 0xffffffff mask flags 0 part_validate: id: TQM8xxL-0 p->off: 00060000 p->sz: 001A0000 end: 200000 id->sz: 00400000 + partition: name rootfs size 0x00200000 offset 0xffffffff mask flags 0 part_validate: id: TQM8xxL-0 p->off: 00200000 p->sz: 00200000 end: 400000 id->sz: 00400000 + partition: name data size 0x00400000 offset 0xffffffff mask flags 0 part_validate: id: TQM8xxL-0 p->off: 00400000 p->sz: 00400000 end: 800000 id->sz: 00400000 TQM8xxL-0: partitioning exceeds flash size It seems that the part_validate() code increases the offset, such as if both banks of NOR flash were indeed just one big device (which I consider correct behaviour), but then it compares it against the size of the current bank instead of against the total flash size. Am I missing something? Best regards, Wolfgang Denk -- DENX Software Engineering GmbH, MD: Wolfgang Denk & Detlev Zundel HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de You'll learn something about men and women -- the way they're supposed to be. Caring for each other, being happy with each other, being good to each other. That's what we call love. You'll like that a lot. -- Kirk, "The Apple", stardate 3715.6