Re: [PATCH 1/2] dt-bindings: mtd: nand: Macronix: document new binding

From: William Zhang
Date: Mon May 22 2023 - 21:01:49 EST


Hi Alvaro,

On 05/17/2023 08:20 AM, Álvaro Fernández Rojas wrote:
Hi William,

El mié, 17 may 2023 a las 7:30, William Zhang
(<william.zhang@xxxxxxxxxxxx>) escribió:



On 05/16/2023 12:02 PM, Álvaro Fernández Rojas wrote:
Sure,

Here you go:
[ 0.000000] Linux version 5.15.111 (noltari@atlantis)
(mips-openwrt-linux-musl-gcc (OpenWrt GCC 12.3.0 r0+22899-466be0612a)
12.3.0, GNU ld (GNU Binutils) 2.40.0) #0 SMP Tue May 16 14:33:20 2023
[ 0.000000] CPU0 revision is: 0002a080 (Broadcom BMIPS4350)
[ 0.000000] MIPS: machine is Sercomm H500-s vfes
[ 0.000000] 128MB of RAM installed
[ 0.000000] earlycon: bcm63xx_uart0 at MMIO 0x10000180 (options '115200n8')
[ 0.000000] printk: bootconsole [bcm63xx_uart0] enabled
[ 0.000000] Initrd not found or empty - disabling initrd
[ 0.000000] Reserving 0KB of memory at 4194303KB for kdump
[ 0.000000] Primary instruction cache 64kB, VIPT, 4-way, linesize 16 bytes.
[ 0.000000] Primary data cache 32kB, 2-way, VIPT, cache aliases,
linesize 16 bytes
[ 0.000000] Zone ranges:
[ 0.000000] Normal [mem 0x0000000000000000-0x0000000007ffffff]
[ 0.000000] Movable zone start for each node
[ 0.000000] Early memory node ranges
[ 0.000000] node 0: [mem 0x0000000000000000-0x0000000007ffffff]
[ 0.000000] Initmem setup node 0 [mem 0x0000000000000000-0x0000000007ffffff]
[ 0.000000] percpu: Embedded 11 pages/cpu s13328 r8192 d23536 u45056
[ 0.000000] Built 1 zonelists, mobility grouping on. Total pages: 32480
[ 0.000000] Kernel command line: earlycon
[ 0.000000] Dentry cache hash table entries: 16384 (order: 4, 65536
bytes, linear)
[ 0.000000] Inode-cache hash table entries: 8192 (order: 3, 32768
bytes, linear)
[ 0.000000] mem auto-init: stack:off, heap alloc:off, heap free:off
[ 0.000000] Memory: 108656K/131072K available (6902K kernel code,
613K rwdata, 1404K rodata, 11872K init, 215K bss, 22416K reserved, 0K
cma-reserved)
[ 0.000000] SLUB: HWalign=64, Order=0-3, MinObjects=0, CPUs=2, Nodes=1
[ 0.000000] rcu: Hierarchical RCU implementation.
[ 0.000000] Tracing variant of Tasks RCU enabled.
[ 0.000000] rcu: RCU calculated value of scheduler-enlistment delay
is 10 jiffies.
[ 0.000000] NR_IRQS: 256
[ 0.000000] irq_bcm6345_l1: registered BCM6345 L1 intc (IRQs: 128)
[ 0.000000] irq_bcm6345_l1: CPU0 (irq = 2)
[ 0.000000] irq_bcm6345_l1: CPU1 (irq = 3)
[ 0.000000] brcm,bcm63268 detected @ 400 MHz
[ 0.000000] clocksource: MIPS: mask: 0xffffffff max_cycles:
0xffffffff, max_idle_ns: 9556302233 ns
[ 0.000002] sched_clock: 32 bits at 200MHz, resolution 5ns, wraps
every 10737418237ns
[ 0.008292] Calibrating delay loop... 398.13 BogoMIPS (lpj=1990656)
[ 0.074683] pid_max: default: 32768 minimum: 301
[ 0.081788] Mount-cache hash table entries: 1024 (order: 0, 4096
bytes, linear)
[ 0.089319] Mountpoint-cache hash table entries: 1024 (order: 0,
4096 bytes, linear)
[ 0.106094] rcu: Hierarchical SRCU implementation.
[ 0.112665] smp: Bringing up secondary CPUs ...
[ 0.119348] SMP: Booting CPU1...
[ 8.330979] Primary instruction cache 64kB, VIPT, 4-way, linesize 16 bytes.
[ 8.331017] Primary data cache 32kB, 2-way, VIPT, cache aliases,
linesize 16 bytes
[ 8.331294] CPU1 revision is: 0002a080 (Broadcom BMIPS4350)
[ 0.182819] Synchronize counters for CPU 1:
[ 0.203500] SMP: CPU1 is running
[ 0.203512] done.
[ 0.213401] smp: Brought up 1 node, 2 CPUs
[ 0.228870] clocksource: jiffies: mask: 0xffffffff max_cycles:
0xffffffff, max_idle_ns: 19112604462750000 ns
[ 0.239058] futex hash table entries: 512 (order: 3, 32768 bytes, linear)
[ 0.246439] pinctrl core: initialized pinctrl subsystem
[ 0.254917] NET: Registered PF_NETLINK/PF_ROUTE protocol family
[ 0.312700] clocksource: Switched to clocksource MIPS
[ 0.321061] NET: Registered PF_INET protocol family
[ 0.326879] IP idents hash table entries: 2048 (order: 2, 16384
bytes, linear)
[ 0.335972] tcp_listen_portaddr_hash hash table entries: 512
(order: 0, 6144 bytes, linear)
[ 0.344721] Table-perturb hash table entries: 65536 (order: 6,
262144 bytes, linear)
[ 0.352721] TCP established hash table entries: 1024 (order: 0,
4096 bytes, linear)
[ 0.360622] TCP bind hash table entries: 1024 (order: 1, 8192 bytes, linear)
[ 0.368005] TCP: Hash tables configured (established 1024 bind 1024)
[ 0.375074] UDP hash table entries: 256 (order: 1, 8192 bytes, linear)
[ 0.381862] UDP-Lite hash table entries: 256 (order: 1, 8192 bytes, linear)
[ 0.389762] NET: Registered PF_UNIX/PF_LOCAL protocol family
[ 0.395748] PCI: CLS 0 bytes, default 16
[ 0.403410] workingset: timestamp_bits=14 max_order=15 bucket_order=1
[ 0.426490] squashfs: version 4.0 (2009/01/31) Phillip Lougher
[ 0.432492] jffs2: version 2.2 (NAND) (SUMMARY) (LZMA) (RTIME)
(CMODE_PRIORITY) (c) 2001-2006 Red Hat, Inc.
[ 0.459472] bcm63xx-power-controller 1000184c.power-controller:
registered 14 power domains
[ 0.470267] 10000180.serial: ttyS0 at MMIO 0x10000180 (irq = 8,
base_baud = 1562500) is a bcm63xx_uart
[ 0.479996] printk: console [ttyS0] enabled
[ 0.479996] printk: console [ttyS0] enabled
[ 0.488651] printk: bootconsole [bcm63xx_uart0] disabled
[ 0.488651] printk: bootconsole [bcm63xx_uart0] disabled
[ 0.533435] bcm2835-rng 10002880.rng: hwrng registered
[ 0.606025] bcm6368_nand 10000200.nand: there is not valid maps for
state default
[ 0.633977] nand: device found, Manufacturer ID: 0xc2, Chip ID: 0xf1
[ 0.640506] nand: Macronix MX30LF1G18AC
[ 0.644551] nand: 128 MiB, SLC, erase size: 128 KiB, page size:
2048, OOB size: 64
[ 0.652359] bcm6368_nand 10000200.nand: detected 128MiB total,
128KiB blocks, 2KiB pages, 16B OOB, 8-bit, BCH-4
[ 0.703373] Bad block table not found for chip 0
[ 0.732040] Bad block table not found for chip 0
[ 0.736842] Scanning device for bad blocks
[ 0.832678] CPU 0 Unable to handle kernel paging request at virtual
address 00000014, epc == 8009b300, ra == 806cc650
[ 0.843628] Oops[#1]:
[ 0.845958] CPU: 0 PID: 88 Comm: hwrng Not tainted 5.15.111 #0
[ 0.851959] $ 0 : 00000000 00000001 00000008 00000000
[ 0.857358] $ 4 : 81808464 00000064 00000000 00000001
[ 0.862753] $ 8 : 81810000 00001ff0 00001c00 815b8880
[ 0.868146] $12 : 0000b79d 00000000 00000000 00009bb

Please, tell me if you want me to add any debugging to the log.

Best regards,
Álvaro.

El mar, 16 may 2023 a las 20:58, Florian Fainelli
(<f.fainelli@xxxxxxxxx>) escribió:

+William,

On 5/16/23 11:55, Álvaro Fernández Rojas wrote:
Hi Jaime,

I've reproduced the issue on a Comtrend VR-3032u (MX30LF1G08AA). After
forcing it to check block protection (it's not supported on that
device), the NAND controller stops reading/writing anything.

@Florian is it possible that low level ops (GET_FEATURES/SET_FEATURES)
aren't supported on BCM63268 NAND controllers and this is causing the
issue?

Yes, this looks like what we have seen as well even with newer NAND
controllers actually. Would it be possible to obtain a full log from
either of you?

William, is this something you have seen before as well?

No, I haven't seen such issue before. It is possible I didn't have this
Macronix parts in my board. If I can find a board with Macronix part,
I will try it. But we don't use this feature and don't connect the PT
pin in our reference board which means the PT feature is disabled in the
nand part.

Alvaro, Do you know if your 63268 board has PT pin connected or not?

No, I don't know if PT pin is connected.
I would have to open the case and check, but judging from the
following image I would say it's not connected:
https://openwrt.org/_media/media/sercomm/h500s/h500s-nand.jpg

Can you check if the macronix's lock and unlock function being calling
before the hang? Or is it just get/set feature function getting called
to determine PT is supported? The get/set feature function should work
as they are used by other pathes

No, the macronix's lock/unlock functions aren't called before the hang.
In fact, if I comment out the nand_get_features call and replace it
with ret = 1 it doesn't hang:
https://github.com/torvalds/linux/blob/f1fcbaa18b28dec10281551dfe6ed3a3ed80e3d6/drivers/mtd/nand/raw/nand_macronix.c#L229-L230

I see. In fact I saw your earlier debug log with ll_op cmd dump for the nand_get_features function and they went through successfully. Really strange how this function call will cause problem to subsequent nand read.

Can you keep this code commented out and then after the board boot up and manually write to nand controller register for these ll_op cmd sequence following the code of brcmnand_low_level_op, assuming you have a way to write to controller registering from a shell. If not, you might have to hack the brcmnand or base nand driver code and insert following call at some special condition at run time:
ret = nand_get_features(chip, ONFI_FEATURE_ADDR_MXIC_PROTECTION,
feature);
Then check if nand read function still works. At least we can confirm if feature query function actually cause the problem. You can try different feature code and see if it make any difference.

Question to Jaime, if PT pin is not connected, would the PT feature check cause any issue afterwards? Or the nand chip should just return block not protected?




Best regards,
Álvaro.

El mié, 26 abr 2023 a las 9:24, liao jaime (<jaimeliao.tw@xxxxxxxxx>) escribió:

Hi Álvaro

In nand_scan_tail(), each manufacturer init function call will be execute.
In macronix_nand_init(), block protect will be execute after flash detect.
I have validate MX30LF1G18AC in Linux kernel v5.15.
I didn't got situation "device hangs" on my side.
BP is to prevent incorrect operations.
Please check the controller settings for tracing this issue.

Thanks
Jaime


Hello YouChing and Jaime,

I still didn't get any feedback from you (or Macronix) on this issue.
Did you have time to look into it?

Thanks,
Álvaro.

El vie, 24 mar 2023 a las 18:04, Álvaro Fernández Rojas
(<noltari@xxxxxxxxx>) escribió:

Hi Miquèl,

2023-03-24 15:36 GMT+01:00, Miquel Raynal <miquel.raynal@xxxxxxxxxxx>:
Hi Álvaro,

+ YouChing and Jaime from Macronix
TLDR for them: there is a misbehavior since Mason added block
protection support. Just checking if the blocks are protected seems to
misconfigure the chip entirely, see below. Any hints?

Could it be that the NAND is stuck expecting a read 0x00 command which
isn’t sent after getting the features?


noltari@xxxxxxxxx wrote on Fri, 24 Mar 2023 15:15:47 +0100:

Hi Miquèl,

2023-03-24 14:45 GMT+01:00, Miquel Raynal <miquel.raynal@xxxxxxxxxxx>:
Hi Álvaro,

noltari@xxxxxxxxx wrote on Fri, 24 Mar 2023 12:21:11 +0100:

El vie, 24 mar 2023 a las 11:49, Miquel Raynal
(<miquel.raynal@xxxxxxxxxxx>) escribió:

Hi Álvaro,

noltari@xxxxxxxxx wrote on Fri, 24 Mar 2023 11:31:17 +0100:

Hi Miquèl,

El vie, 24 mar 2023 a las 10:40, Miquel Raynal
(<miquel.raynal@xxxxxxxxxxx>) escribió:

Hi Álvaro,

noltari@xxxxxxxxx wrote on Thu, 23 Mar 2023 13:45:09 +0100:

Add new "mxic,disable-block-protection" binding documentation.
This binding allows disabling block protection support for
those
devices not
supporting it.

Signed-off-by: Álvaro Fernández Rojas <noltari@xxxxxxxxx>
---
Documentation/devicetree/bindings/mtd/nand-macronix.txt | 3
+++
1 file changed, 3 insertions(+)

diff --git
a/Documentation/devicetree/bindings/mtd/nand-macronix.txt
b/Documentation/devicetree/bindings/mtd/nand-macronix.txt
index ffab28a2c4d1..03f65ca32cd3 100644
--- a/Documentation/devicetree/bindings/mtd/nand-macronix.txt
+++ b/Documentation/devicetree/bindings/mtd/nand-macronix.txt
@@ -16,6 +16,9 @@ in children nodes.
Required NAND chip properties in children mode:
- randomizer enable: should be "mxic,enable-randomizer-otp"

+Optional NAND chip properties in children mode:
+- block protection disable: should be
"mxic,disable-block-protection"
+

Besides the fact that nowadays we prefer to see binding
conversions
to
yaml before adding anything, I don't think this will fly.

I'm not sure exactly what "disable block protection" means, we
already have similar properties like "lock" and
"secure-regions",
not
sure they will fit but I think it's worth checking.

As explained in 2/2, commit 03a539c7a118 introduced a regression
on
Sercomm H500-s (BCM63268) OpenWrt devices with Macronix
MX30LF1G18AC
which hangs the device.

This is the log with block protection disabled:
[ 0.495831] bcm6368_nand 10000200.nand: there is not valid maps
for
state default
[ 0.504995] nand: device found, Manufacturer ID: 0xc2, Chip ID:
0xf1
[ 0.511526] nand: Macronix MX30LF1G18AC
[ 0.515586] nand: 128 MiB, SLC, erase size: 128 KiB, page size:
2048, OOB size: 64
[ 0.523516] bcm6368_nand 10000200.nand: detected 128MiB total,
128KiB blocks, 2KiB pages, 16B OOB, 8-bit, BCH-4
[ 0.535912] Bad block table found at page 65472, version 0x01
[ 0.544268] Bad block table found at page 65408, version 0x01
[ 0.954329] 9 fixed-partitions partitions found on MTD device
brcmnand.0
...

This is the log with block protection enabled:
[ 0.495095] bcm6368_nand 10000200.nand: there is not valid maps
for
state default
[ 0.504249] nand: device found, Manufacturer ID: 0xc2, Chip ID:
0xf1
[ 0.510772] nand: Macronix MX30LF1G18AC
[ 0.514874] nand: 128 MiB, SLC, erase size: 128 KiB, page size:
2048, OOB size: 64
[ 0.522780] bcm6368_nand 10000200.nand: detected 128MiB total,
128KiB blocks, 2KiB pages, 16B OOB, 8-bit, BCH-4
[ 0.539687] Bad block table not found for chip 0
[ 0.550153] Bad block table not found for chip 0
[ 0.555069] Scanning device for bad blocks
[ 0.601213] CPU 1 Unable to handle kernel paging request at
virtual
address 10277f00, epc == 8039ce70, ra == 8016ad50
*** Device hangs ***

Enabling macronix_nand_block_protection_support() makes the device
unable to detect the bad block table and hangs it when trying to
scan
for bad blocks.

Please trace nand_macronix.c and look:
- are the get_features and set_features really supported by the
controller driver?

This is what I could find by debugging:
[ 0.494993] bcm6368_nand 10000200.nand: there is not valid maps for
state default
[ 0.505375] nand: device found, Manufacturer ID: 0xc2, Chip ID:
0xf1
[ 0.512077] nand: Macronix MX30LF1G18AC
[ 0.515994] nand: 128 MiB, SLC, erase size: 128 KiB, page size:
2048, OOB size: 64
[ 0.523928] bcm6368_nand 10000200.nand: detected 128MiB total,
128KiB blocks, 2KiB pages, 16B OOB, 8-bit, BCH-4
[ 0.534415] bcm6368_nand 10000200.nand: ll_op cmd 0xa00ee
[ 0.539988] bcm6368_nand 10000200.nand: ll_op cmd 0x600a0
[ 0.545659] bcm6368_nand 10000200.nand: ll_op cmd 0x10000
[ 0.551214] bcm6368_nand 10000200.nand: NAND_CMD_GET_FEATURES =
0x00
[ 0.557843] bcm6368_nand 10000200.nand: ll_op cmd 0x10000
[ 0.563475] bcm6368_nand 10000200.nand: NAND_CMD_GET_FEATURES =
0x00
[ 0.569998] bcm6368_nand 10000200.nand: ll_op cmd 0x10000
[ 0.575653] bcm6368_nand 10000200.nand: NAND_CMD_GET_FEATURES =
0x00
[ 0.582246] bcm6368_nand 10000200.nand: ll_op cmd 0x80010000
[ 0.588067] bcm6368_nand 10000200.nand: NAND_CMD_GET_FEATURES =
0x00
[ 0.594657] nand: nand_get_features: addr=a0 subfeature_param=[00
00 00 00] -> 0
[ 0.602341] macronix_nand_block_protection_support:
ONFI_FEATURE_ADDR_MXIC_PROTECTION=0
[ 0.610548] macronix_nand_block_protection_support: !=
MXIC_BLOCK_PROTECTION_ALL_LOCK
[ 0.624760] Bad block table not found for chip 0
[ 0.635542] Bad block table not found for chip 0
[ 0.640270] Scanning device for bad blocks

I don't know how to tell if get_features / set_features is really
supported...

Looks like your driver does not support exec_op but the core provides a
get/set_feature implementation.

According to Florian, low level should be supported on brcmnand
controllers >= 4.0
Also:
https://github.com/nomis/bcm963xx_4.12L.06B_consumer/blob/e2f23ddbb20bf75689372b6e6a5a0dc613f6e313/shared/opensource/include/bcm963xx/63268_map_part.h#L1597

Just to be sure, you're using a mainline controller driver, not this
one?

Yes, this was just to prove that the HW I’m using has get/set features support.
I’m using OpenWrt, so it’s linux v5.15 driver.




- what is the state of the locking configuration in the chip when
you
boot?

Unlocked, I guess...
How can I check that?

It's in your dump, the chip returns 0, meaning it's all unlocked,
apparently.

Well, I can read/write the device if block protection isn’t disabled,
so I guess we can confirm it’s unlocked…


- is there anything that locks the device by calling mxic_nand_lock()
?

So nobody locks the device I guess? Did you add traces there?

It doesn’t get to the point that it enabled the lock/unlock functions
since it fails when checking if feature is 0x38, so there’s no point
in adding those traces…

Right, it returns before setting these I guess.



- finding no bbt is one thing, hanging is another, where is it
hanging
exactly? (offset in nand/ and line in the code)

I've got no idea...

You can use ftrace or just add printks a bit everywhere and try to get
closer and closer.

I think that after trying to get the feature it just start reading
nonsense from the NAND and at some point it hangs due to that garbage…

It should refuse to mount the device somehow, but in no case the kernel
should hang.

Yes, I think that this is a side effect (maybe a different bug somewhere else).


Is it posible that the NAND starts behaving like this after getting
the feature due to some specific config of my device?


I looked at the patch, I don't see anything strange. Besides, I have a
close enough datasheet and I don't see what could confuse the device.

Are you really sure this patch is the problem? Is the WP pin wired on
your design?

There’s no WP pin in brcmnand controllers < 7.0

What about the chip?

Maybe it has a GPIO controlling that, but I don’t have that info…


Thanks,
Miquèl


--
Florian


--
Álvaro

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature