| Message ID | 20230125081023.1573712-5-d-gole@ti.com |
|---|---|
| State | Handled Elsewhere |
| Headers | show |
| Series | STIG Mode Fixes for spi-cadence-qspi driver | expand |
Hi Dhruva, An error occurred in the following environments where this patch was applied, and the flash could not be accessed. However, after reverting to this patch, it is possible to access it. - Environment - OS : Linux 6.3 - SoC : Cyclone V - Flash : MT25QL512A - Error at startup - Kernel log [ 0.980290] spi-nor spi0.0: found mt25ql512a, expected n25q512a [ 1.485140] cadence-qspi ff705000.flash: Flash command execution timed out. [ 1.490792] spi-nor spi0.0: operation failed with -110 [ 1.494654] spi-nor spi0.0: mt25ql512a (65536 Kbytes) - Error at access - Access command and log # hexdump -Cv /dev/mtdblock0 hexdump: /dev/mtdblock0: Input/output error - Kernel log [ 2124.201193] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. [ 2124.201229] spi-nor spi0.0: operation failed with -110 [ 2124.201256] I/O error, dev mtdblock0, sector 0 op 0x0:(READ) flags 0x80700 phys_seg 4 prio class 2 [ 2124.711241] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. [ 2124.711276] spi-nor spi0.0: operation failed with -110 [ 2124.711302] I/O error, dev mtdblock0, sector 8 op 0x0:(READ) flags 0x80700 phys_seg 3 prio class 2 [ 2125.221193] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. [ 2125.221230] spi-nor spi0.0: operation failed with -110 [ 2125.221256] I/O error, dev mtdblock0, sector 16 op 0x0:(READ) flags 0x80700 phys_seg 2 prio class 2 [ 2125.731237] cadence-qspi ff705000.flash:. QSPI is still busy after 500ms timeout. [ 2125.731270] spi-nor spi0.0: operation failed with -110 [ 2125.731296] I/O error, dev mtdblock0, sector 24 op 0x0:(READ) flags 0x80700 phys_seg 1 prio class 2 [ 2126.241190] cadence-qspi ff705000.flash:. QSPI is still busy after 500ms timeout. [ 2126.241224] spi-nor spi0.0: operation failed with -110 [ 2126.241251] I/O error, dev mtdblock0, sector 0 op 0x0:(READ) flags 0x0 phys_seg 1 prio class 2 [ 2126.241274] Buffer I/O error on dev mtdblock0, logical block 0, async page read regards, Yoshitaka Ikeda -----Original Message----- From: Dhruva Gole <d-gole@ti.com> Sent: Wednesday, January 25, 2023 5:10 PM To: Mark Brown <broonie@kernel.org> Cc: linux-spi@vger.kernel.org; linux-kernel@vger.kernel.org; Vignesh Raghavendra <vigneshr@ti.com>; Vaishnav Achath <vaishnav.a@ti.com>; Dhruva Gole <d-gole@ti.com>; linux-mtd@lists.infradead.org; linux-arm-kernel@lists.infradead.org; Takahiro.Kuwano@infineon.com; Pratyush Yadav <ptyadav@amazon.de> Subject: [PATCH v2 4/4] spi: cadence-quadspi: use STIG mode for small reads Fix the issue where some flash chips like cypress S25HS256T return the value of the same register over and over in DAC mode. For example in the TI K3-AM62x Processors refer [0] Technical Reference Manual there is a layer of digital logic in front of the QSPI/OSPI Drive when used in DAC mode. This is part of the Flash Subsystem (FSS) which provides access to external Flash devices. The FSS0_0_SYSCONFIG Register (Offset = 4h) has a BIT Field for OSPI_32B_DISABLE_MODE which has a Reset value = 0. This means, OSPI 32bit mode enabled by default. Thus, by default controller operates in 32 bit mode causing it to always align all data to 4 bytes from a 4byte aligned address. In some flash chips like cypress for example if we try to read some regs in DAC mode then it keeps sending the value of the first register that was requested and inorder to read the next reg, we have to stop and re-initiate a new transaction. This causes wrong register values to be read than what is desired when registers are read in DAC mode. Hence if the data.nbytes is very less then prefer STIG mode for such small reads. [0] https://www.ti.com/lit/ug/spruiv7a/spruiv7a.pdf Signed-off-by: Dhruva Gole <d-gole@ti.com> --- drivers/spi/spi-cadence-quadspi.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) diff --git a/drivers/spi/spi-cadence-quadspi.c b/drivers/spi/spi-cadence-quadspi.c index a6d0f1b0bb49..2954c06a7f57 100644 --- a/drivers/spi/spi-cadence-quadspi.c +++ b/drivers/spi/spi-cadence-quadspi.c @@ -1359,7 +1359,13 @@ static int cqspi_mem_process(struct spi_mem *mem, const struct spi_mem_op *op) cqspi_configure(f_pdata, mem->spi->max_speed_hz); if (op->data.dir == SPI_MEM_DATA_IN && op->data.buf.in) { - if (!op->addr.nbytes) + /* + * Performing reads in DAC mode forces to read minimum 4 bytes + * which is unsupported on some flash devices during register + * reads, prefer STIG mode for such small reads. + */ + if (!op->addr.nbytes || + op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX) return cqspi_command_read(f_pdata, op); return cqspi_read(f_pdata, op); -- 2.25.1
Hi Yoshitaka, On 26/04/23 08:04, Yoshitaka Ikeda wrote: > Hi Dhruva, > > An error occurred in the following environments where this patch was applied, and the flash could not be accessed. > However, after reverting to this patch, it is possible to access it. > > - Environment > - OS : Linux 6.3 > - SoC : Cyclone V Please can you send me the register fields information for the CQSPI controller used in this device? I wanted to verify if atall there were any mismatch between the controller I have tested with vs your SOC's controller. > - Flash : MT25QL512A > > - Error at startup > - Kernel log > [ 0.980290] spi-nor spi0.0: found mt25ql512a, expected n25q512a > [ 1.485140] cadence-qspi ff705000.flash: Flash command execution timed out. > [ 1.490792] spi-nor spi0.0: operation failed with -110 > [ 1.494654] spi-nor spi0.0: mt25ql512a (65536 Kbytes) > > - Error at access > - Access command and log > # hexdump -Cv /dev/mtdblock0 > hexdump: /dev/mtdblock0: Input/output error > > - Kernel log > [ 2124.201193] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. > [ 2124.201229] spi-nor spi0.0: operation failed with -110 > [ 2124.201256] I/O error, dev mtdblock0, sector 0 op 0x0:(READ) flags 0x80700 phys_seg 4 prio class 2 > [ 2124.711241] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. > [ 2124.711276] spi-nor spi0.0: operation failed with -110 > [ 2124.711302] I/O error, dev mtdblock0, sector 8 op 0x0:(READ) flags 0x80700 phys_seg 3 prio class 2 > [ 2125.221193] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. > [ 2125.221230] spi-nor spi0.0: operation failed with -110 > [ 2125.221256] I/O error, dev mtdblock0, sector 16 op 0x0:(READ) flags 0x80700 phys_seg 2 prio class 2 > [ 2125.731237] cadence-qspi ff705000.flash:. QSPI is still busy after 500ms timeout. > [ 2125.731270] spi-nor spi0.0: operation failed with -110 > [ 2125.731296] I/O error, dev mtdblock0, sector 24 op 0x0:(READ) flags 0x80700 phys_seg 1 prio class 2 > [ 2126.241190] cadence-qspi ff705000.flash:. QSPI is still busy after 500ms timeout. > [ 2126.241224] spi-nor spi0.0: operation failed with -110 > [ 2126.241251] I/O error, dev mtdblock0, sector 0 op 0x0:(READ) flags 0x0 phys_seg 1 prio class 2 > [ 2126.241274] Buffer I/O error on dev mtdblock0, logical block 0, async page read > > > regards, > Yoshitaka Ikeda > > > [..snip..] It maybe that STIG mode may have some different way of configuring/using on your SoC. However at this point it's only a guess and I will wait till I have the exact information on the controller specs being used in this device along with the register descriptions.
Hi Dhruva, > Please can you send me the register fields information for the CQSPI > controller used in this device? > I wanted to verify if atall there were any mismatch between the controller I > have tested with vs your SOC's controller. You can find it in the pdf that can be obtained by pressing the download button in the upper right corner of the following website. - https://www.intel.com/content/www/us/en/docs/programmable/683126/15-0/introduction.html The description is as follows: - 15. Quad SPI Flash - Quad SPI Flash Controller Address Map and Register Definitions
Hi, On 27/04/23 06:11, Yoshitaka Ikeda wrote: > Hi Dhruva, > >> Please can you send me the register fields information for the CQSPI >> controller used in this device? >> I wanted to verify if atall there were any mismatch between the controller I >> have tested with vs your SOC's controller. > > You can find it in the pdf that can be obtained by pressing the download button in the upper right corner of the following website. > - https://www.intel.com/content/www/us/en/docs/programmable/683126/15-0/introduction.html Thanks for sharing, I went through and couldn't really find any major changes at the controller level. So I am wondering if some specific sequence of flash operations on your device is exposing some issues in the driver's STIG reads. Please can you share some logs with the following patch: I am trying to see a pattern that may be causing issues. I am unable to reproduce this on my end at the moment. ... diff --git a/drivers/spi/spi-cadence-quadspi.c b/drivers/spi/spi-cadence-quadspi.c index 6ddb2dfc0f00..5fa71dbd37ef 100644 --- a/drivers/spi/spi-cadence-quadspi.c +++ b/drivers/spi/spi-cadence-quadspi.c @@ -1390,6 +1390,14 @@ static int cqspi_exec_mem_op(struct spi_mem *mem, const struct spi_mem_op *op) int ret; ret = cqspi_mem_process(mem, op); + /* debug */ + printk("**********spi_mem_op dump**************"); + printk("addr: nbytes:0x%X , buswidth 0x%X, dtr 0x%X, val 0x%llX", op->addr.nbytes, op->addr.buswidth, op->addr.dtr, op->addr.val); + printk("cmd: nbytes:0x%X , buswidth 0x%X, dtr 0x%X, opcode 0x%X", op->cmd.nbytes, op->cmd.buswidth, op->cmd.dtr, op->cmd.opcode); + printk("data: nbytes:0x%X , buswidth 0x%X, dtr 0x%X, data dir 0x%X", op->data.nbytes, op->data.buswidth, op->data.dtr, op->data.dir); + + printk("***************************************"); + /* ---------------------------- */ if (ret) dev_err(&mem->spi->dev, "operation failed with %d\n", ret); ...
Hi Dhruva, Sorry for the late reply as I was on a long vacation. > Thanks for sharing, I went through and couldn't really find any major changes > at the controller level. > So I am wondering if some specific sequence of flash operations on your > device is exposing some issues in the driver's STIG reads. Thank you for looking into this. > Please can you share some logs with the following patch: > I am trying to see a pattern that may be causing issues. > I am unable to reproduce this on my end at the moment. The logs obtained with the patch applied are as follows: - Error at startup - Kernel log [ 0.980598] **********spi_mem_op dump************** [ 0.980613] addr: nbytes:0x0 , buswidth 0x0, dtr 0x0, val 0x0 [ 0.984223] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x9F [ 0.988656] data: nbytes:0x6 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 0.993362] *************************************** [ 0.998329] spi-nor spi0.0: found mt25ql512a, expected n25q512a [ 1.006574] **********spi_mem_op dump************** [ 1.006583] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x0 [ 1.010150] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A [ 1.014596] data: nbytes:0x10 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 1.019285] *************************************** [ 1.524271] cadence-qspi ff705000.flash: Flash command execution timed out. [ 1.533483] **********spi_mem_op dump************** [ 1.533489] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x10 [ 1.537055] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A [ 1.541579] data: nbytes:0x8 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 1.546266] *************************************** [ 1.551123] spi-nor spi0.0: operation failed with -110 [ 1.558531] spi-nor spi0.0: mt25ql512a (65536 Kbytes) - Error at access - Access command and log # hexdump -Cv /dev/mtdblock0 hexdump: /dev/mtdblock0: Input/output error - Kernel log [ 75.401297] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. [ 75.401337] **********spi_mem_op dump************** [ 75.401343] addr: nbytes:0x4 , buswidth 0x1, dtr 0x0, val 0x0 [ 75.401354] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0xC [ 75.401364] data: nbytes:0x200 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 75.401373] *************************************** [ 75.401382] spi-nor spi0.0: operation failed with -110 [ 75.401407] I/O error, dev mtdblock0, sector 0 op 0x0:(READ) flags 0x80700 phys_seg 4 prio class 2 [ 75.911339] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. [ 75.911372] **********spi_mem_op dump************** [ 75.911379] addr: nbytes:0x4 , buswidth 0x1, dtr 0x0, val 0x1000 [ 75.911390] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0xC [ 75.911399] data: nbytes:0x200 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 75.911409] *************************************** [ 75.911417] spi-nor spi0.0: operation failed with -110 [ 75.911443] I/O error, dev mtdblock0, sector 8 op 0x0:(READ) flags 0x80700 phys_seg 3 prio class 2 [ 76.421333] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. [ 76.421365] **********spi_mem_op dump************** [ 76.421371] addr: nbytes:0x4 , buswidth 0x1, dtr 0x0, val 0x2000 [ 76.421383] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0xC [ 76.421393] data: nbytes:0x200 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 76.421403] *************************************** [ 76.421411] spi-nor spi0.0: operation failed with -110 [ 76.421436] I/O error, dev mtdblock0, sector 16 op 0x0:(READ) flags 0x80700 phys_seg 2 prio class 2 [ 76.931330] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. [ 76.931365] **********spi_mem_op dump************** [ 76.931371] addr: nbytes:0x4 , buswidth 0x1, dtr 0x0, val 0x3000 [ 76.931382] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0xC [ 76.931392] data: nbytes:0x200 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 76.931401] *************************************** [ 76.931410] spi-nor spi0.0: operation failed with -110 [ 76.931437] I/O error, dev mtdblock0, sector 24 op 0x0:(READ) flags 0x80700 phys_seg 1 prio class 2 [ 77.441297] cadence-qspi ff705000.flash: QSPI is still busy after 500ms timeout. [ 77.441330] **********spi_mem_op dump************** [ 77.441337] addr: nbytes:0x4 , buswidth 0x1, dtr 0x0, val 0x0 [ 77.441347] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0xC [ 77.441357] data: nbytes:0x200 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 77.441367] *************************************** [ 77.441375] spi-nor spi0.0: operation failed with -110 [ 77.441401] I/O error, dev mtdblock0, sector 0 op 0x0:(READ) flags 0x0 phys_seg 1 prio class 2 [ 77.441423] Buffer I/O error on dev mtdblock0, logical block 0, async page read
On 08/05/23 06:06, Yoshitaka Ikeda wrote: > Hi Dhruva, > > Sorry for the late reply as I was on a long vacation. > >> Thanks for sharing, I went through and couldn't really find any major changes >> at the controller level. >> So I am wondering if some specific sequence of flash operations on your >> device is exposing some issues in the driver's STIG reads. > > Thank you for looking into this. > >> Please can you share some logs with the following patch: >> I am trying to see a pattern that may be causing issues. >> I am unable to reproduce this on my end at the moment. > > The logs obtained with the patch applied are as follows: Thanks for sharing the logs. It has made me concerned about something, > > - Error at startup > - Kernel log > [ 0.980598] **********spi_mem_op dump************** > [ 0.980613] addr: nbytes:0x0 , buswidth 0x0, dtr 0x0, val 0x0 > [ 0.984223] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x9F > [ 0.988656] data: nbytes:0x6 , buswidth 0x1, dtr 0x0, data dir 0x1 > [ 0.993362] *************************************** > [ 0.998329] spi-nor spi0.0: found mt25ql512a, expected n25q512a > [ 1.006574] **********spi_mem_op dump************** > [ 1.006583] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x0 > [ 1.010150] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A > [ 1.014596] data: nbytes:0x10 , buswidth 0x1, dtr 0x0, data dir 0x1 > [ 1.019285] *************************************** > [ 1.524271] cadence-qspi ff705000.flash: Flash command execution timed out. This print message is from cqspi_exec_flash_cmd. This function should only be called from cqspi_command_read/write . However, from spi_mem_op dump that you have provided above, where addr.nbytes is 3 and data.nbytes is 0x10 (which is > 8) it should never have entered the cqspi_command_read function. > [ 1.533483] **********spi_mem_op dump************** > [ 1.533489] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x10 > [ 1.537055] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A > [ 1.541579] data: nbytes:0x8 , buswidth 0x1, dtr 0x0, data dir 0x1 > [ 1.546266] *************************************** > [ 1.551123] spi-nor spi0.0: operation failed with -110 > [ 1.558531] spi-nor spi0.0: mt25ql512a (65536 Kbytes) > Anything after the Flash command execution timed out step seems irrelevant because at that point the controller is in some weird state that it never again comes out of. Please can you share the exact output of uname -a where you observe this error? I am now wondering why atall would the following conditions be satisfied in your case: if (!op->addr.nbytes || op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX) either addr.nbytes needs to be 0, which it isn't. or data has to be <= 8 (which again doesn't seem to be the case here) So it should never enter the above "if" chunk. My logs for reference: https://gist.github.com/DhruvaG2000/7185a84de5757e4988f93478f6b75289 Are you carrying any sort of local patches? Can you make sure that the CQSPI_STIG_DATA_LEN_MAX is 8 in your case too? Any other delta from linux-next that I am using in the logs I gave above?
Hi Dhruva, > > - Error at startup > > - Kernel log > > [ 0.980598] **********spi_mem_op dump************** > > [ 0.980613] addr: nbytes:0x0 , buswidth 0x0, dtr 0x0, val 0x0 > > [ 0.984223] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x9F > > [ 0.988656] data: nbytes:0x6 , buswidth 0x1, dtr 0x0, data dir 0x1 > > [ 0.993362] *************************************** > > [ 0.998329] spi-nor spi0.0: found mt25ql512a, expected n25q512a > > [ 1.006574] **********spi_mem_op dump************** > > [ 1.006583] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x0 > > [ 1.010150] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A > > [ 1.014596] data: nbytes:0x10 , buswidth 0x1, dtr 0x0, data dir 0x1 > > [ 1.019285] *************************************** > > [ 1.524271] cadence-qspi ff705000.flash: Flash command execution > timed out. > > This print message is from cqspi_exec_flash_cmd. This function should only > be called from cqspi_command_read/write . > > However, from spi_mem_op dump that you have provided above, where > addr.nbytes is 3 and data.nbytes is 0x10 (which is > 8) it should never have > entered the cqspi_command_read function. The location of the log output is after the call to cqspi_mem_process(), so I assume it is called with the following data. [ 1.533483] **********spi_mem_op dump************** [ 1.533489] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x10 [ 1.537055] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A [ 1.541579] data: nbytes:0x8 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 1.546266] *************************************** Thus, the condition is met and cqspi_command_read() is called. We have also applied the following patch and confirmed that it is called. - patch ``` diff --git a/drivers/spi/spi-cadence-quadspi.c b/drivers/spi/spi-cadence-quadspi.c index 64b6a460d739..8d3681e1f35c 100644 --- a/drivers/spi/spi-cadence-quadspi.c +++ b/drivers/spi/spi-cadence-quadspi.c @@ -1365,15 +1365,21 @@ static int cqspi_mem_process(struct spi_mem *mem, const struct spi_mem_op *op) * reads, prefer STIG mode for such small reads. */ if (!op->addr.nbytes || - op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX) + op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX) { + printk("----- %d:%s() (1) -----\n", __LINE__, __func__); return cqspi_command_read(f_pdata, op); + } + printk("----- %d:%s() (2) -----\n", __LINE__, __func__); return cqspi_read(f_pdata, op); } - if (!op->addr.nbytes || !op->data.buf.out) + if (!op->addr.nbytes || !op->data.buf.out) { + printk("----- %d:%s() (3) -----\n", __LINE__, __func__); return cqspi_command_write(f_pdata, op); + } + printk("----- %d:%s() (4) -----\n", __LINE__, __func__); return cqspi_write(f_pdata, op); } ``` - log ``` [ 0.786600] ----- 1369:cqspi_mem_process() (1) ----- [ 0.790312] **********spi_mem_op dump************** [ 0.790319] addr: nbytes:0x0 , buswidth 0x0, dtr 0x0, val 0x0 [ 0.793885] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x9F [ 0.798325] data: nbytes:0x6 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 0.803013] *************************************** [ 0.807892] spi-nor spi0.0: found mt25ql512a, expected n25q512a [ 0.816062] ----- 1373:cqspi_mem_process() (2) ----- [ 0.819775] **********spi_mem_op dump************** [ 0.819782] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x0 [ 0.823348] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A [ 0.827792] data: nbytes:0x10 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 0.832480] *************************************** [ 0.837442] ----- 1369:cqspi_mem_process() (1) ----- [ 1.344671] cadence-qspi ff705000.flash: Flash command execution timed out. [ 1.350328] **********spi_mem_op dump************** [ 1.350334] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x10 [ 1.353901] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A [ 1.358427] data: nbytes:0x8 , buswidth 0x1, dtr 0x0, data dir 0x1 [ 1.363114] *************************************** [ 1.367981] spi-nor spi0.0: operation failed with -110 [ 1.375376] spi-nor spi0.0: mt25ql512a (65536 Kbytes) ``` > Please can you share the exact output of uname -a where you observe this > error? My environment has shortened the output of uname -a, so it may not be useful. Instead, I will give you information about the kernel I am using. It is the following (v6.3): https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=457391b0380335d5e9a5babdec90ac53928b23b4 > Are you carrying any sort of local patches? Can you make sure that the > CQSPI_STIG_DATA_LEN_MAX is 8 in your case too? There is no local patch and CQSPI_STIG_DATA_LEN_MAX remains 8.
Hi Dhruva, The patch described below was able to avoid this problem. However, we could not find any material to determine if it was correct. Dear all, Is there anyone else in a similar environment? ``` diff --git a/drivers/spi/spi-cadence-quadspi.c b/drivers/spi/spi-cadence-quadspi.c index 8f36e1306e16..4a9535e6c86a 100644 --- a/drivers/spi/spi-cadence-quadspi.c +++ b/drivers/spi/spi-cadence-quadspi.c @@ -1365,7 +1365,9 @@ static int cqspi_mem_process(struct spi_mem *mem, const struct spi_mem_op *op) * reads, prefer STIG mode for such small reads. */ if (!op->addr.nbytes || - op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX) + (op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX && + !of_device_is_compatible(cqspi->pdev->dev.of_node, + "intel,socfpga-qspi"))) return cqspi_command_read(f_pdata, op); return cqspi_read(f_pdata, op); ``` Thanks and Regards, Yoshitaka Ikeda > -----Original Message----- > From: Yoshitaka Ikeda <ikeda@nskint.co.jp> > Sent: Monday, May 8, 2023 4:44 PM > To: Dhruva Gole <d-gole@ti.com> > Cc: linux-spi@vger.kernel.org; linux-kernel@vger.kernel.org; Vignesh > Raghavendra <vigneshr@ti.com>; Vaishnav Achath <vaishnav.a@ti.com>; > linux-mtd@lists.infradead.org; linux-arm-kernel@lists.infradead.org; > Takahiro.Kuwano@infineon.com; Pratyush Yadav <ptyadav@amazon.de>; > Mark Brown <broonie@kernel.org> > Subject: RE: [PATCH v2 4/4] spi: cadence-quadspi: use STIG mode for small > reads > > Hi Dhruva, > > > > - Error at startup > > > - Kernel log > > > [ 0.980598] **********spi_mem_op dump************** > > > [ 0.980613] addr: nbytes:0x0 , buswidth 0x0, dtr 0x0, val 0x0 > > > [ 0.984223] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x9F > > > [ 0.988656] data: nbytes:0x6 , buswidth 0x1, dtr 0x0, data dir 0x1 > > > [ 0.993362] *************************************** > > > [ 0.998329] spi-nor spi0.0: found mt25ql512a, expected n25q512a > > > [ 1.006574] **********spi_mem_op dump************** > > > [ 1.006583] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x0 > > > [ 1.010150] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A > > > [ 1.014596] data: nbytes:0x10 , buswidth 0x1, dtr 0x0, data dir 0x1 > > > [ 1.019285] *************************************** > > > [ 1.524271] cadence-qspi ff705000.flash: Flash command execution > > timed out. > > > > This print message is from cqspi_exec_flash_cmd. This function should > > only be called from cqspi_command_read/write . > > > > However, from spi_mem_op dump that you have provided above, where > > addr.nbytes is 3 and data.nbytes is 0x10 (which is > 8) it should > > never have entered the cqspi_command_read function. > > The location of the log output is after the call to cqspi_mem_process(), so I > assume it is called with the following data. > > [ 1.533483] **********spi_mem_op dump************** [ 1.533489] addr: > nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x10 [ 1.537055] cmd: nbytes:0x1 , > buswidth 0x1, dtr 0x0, opcode 0x5A [ 1.541579] data: nbytes:0x8 , buswidth 0x1, > dtr 0x0, data dir 0x1 [ 1.546266] *************************************** > > Thus, the condition is met and cqspi_command_read() is called. > We have also applied the following patch and confirmed that it is called. > > - patch > ``` > diff --git a/drivers/spi/spi-cadence-quadspi.c > b/drivers/spi/spi-cadence-quadspi.c > index 64b6a460d739..8d3681e1f35c 100644 > --- a/drivers/spi/spi-cadence-quadspi.c > +++ b/drivers/spi/spi-cadence-quadspi.c > @@ -1365,15 +1365,21 @@ static int cqspi_mem_process(struct spi_mem > *mem, const struct spi_mem_op *op) > * reads, prefer STIG mode for such small reads. > */ > if (!op->addr.nbytes || > - op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX) > + op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX) { > + printk("----- %d:%s() (1) -----\n", __LINE__, > __func__); > return cqspi_command_read(f_pdata, op); > + } > > + printk("----- %d:%s() (2) -----\n", __LINE__, __func__); > return cqspi_read(f_pdata, op); > } > > - if (!op->addr.nbytes || !op->data.buf.out) > + if (!op->addr.nbytes || !op->data.buf.out) { > + printk("----- %d:%s() (3) -----\n", __LINE__, __func__); > return cqspi_command_write(f_pdata, op); > + } > > + printk("----- %d:%s() (4) -----\n", __LINE__, __func__); > return cqspi_write(f_pdata, op); > } > > ``` > > - log > ``` > [ 0.786600] ----- 1369:cqspi_mem_process() (1) ----- > [ 0.790312] **********spi_mem_op dump************** > [ 0.790319] addr: nbytes:0x0 , buswidth 0x0, dtr 0x0, val 0x0 > [ 0.793885] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x9F > [ 0.798325] data: nbytes:0x6 , buswidth 0x1, dtr 0x0, data dir 0x1 > [ 0.803013] *************************************** > [ 0.807892] spi-nor spi0.0: found mt25ql512a, expected n25q512a > [ 0.816062] ----- 1373:cqspi_mem_process() (2) ----- > [ 0.819775] **********spi_mem_op dump************** > [ 0.819782] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x0 > [ 0.823348] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A > [ 0.827792] data: nbytes:0x10 , buswidth 0x1, dtr 0x0, data dir 0x1 > [ 0.832480] *************************************** > [ 0.837442] ----- 1369:cqspi_mem_process() (1) ----- > [ 1.344671] cadence-qspi ff705000.flash: Flash command execution timed > out. > [ 1.350328] **********spi_mem_op dump************** > [ 1.350334] addr: nbytes:0x3 , buswidth 0x1, dtr 0x0, val 0x10 > [ 1.353901] cmd: nbytes:0x1 , buswidth 0x1, dtr 0x0, opcode 0x5A > [ 1.358427] data: nbytes:0x8 , buswidth 0x1, dtr 0x0, data dir 0x1 > [ 1.363114] *************************************** > [ 1.367981] spi-nor spi0.0: operation failed with -110 > [ 1.375376] spi-nor spi0.0: mt25ql512a (65536 Kbytes) > ``` > > > Please can you share the exact output of uname -a where you observe > > this error? > > My environment has shortened the output of uname -a, so it may not be useful. > Instead, I will give you information about the kernel I am using. > It is the following (v6.3): > https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id > =457391b0380335d5e9a5babdec90ac53928b23b4 > > > Are you carrying any sort of local patches? Can you make sure that the > > CQSPI_STIG_DATA_LEN_MAX is 8 in your case too? > > There is no local patch and CQSPI_STIG_DATA_LEN_MAX remains 8. > > -- > Thanks and Regards, > Yoshitaka Ikeda
diff --git a/drivers/spi/spi-cadence-quadspi.c b/drivers/spi/spi-cadence-quadspi.c index a6d0f1b0bb49..2954c06a7f57 100644 --- a/drivers/spi/spi-cadence-quadspi.c +++ b/drivers/spi/spi-cadence-quadspi.c @@ -1359,7 +1359,13 @@ static int cqspi_mem_process(struct spi_mem *mem, const struct spi_mem_op *op) cqspi_configure(f_pdata, mem->spi->max_speed_hz); if (op->data.dir == SPI_MEM_DATA_IN && op->data.buf.in) { - if (!op->addr.nbytes) + /* + * Performing reads in DAC mode forces to read minimum 4 bytes + * which is unsupported on some flash devices during register + * reads, prefer STIG mode for such small reads. + */ + if (!op->addr.nbytes || + op->data.nbytes <= CQSPI_STIG_DATA_LEN_MAX) return cqspi_command_read(f_pdata, op); return cqspi_read(f_pdata, op);
Fix the issue where some flash chips like cypress S25HS256T return the value of the same register over and over in DAC mode. For example in the TI K3-AM62x Processors refer [0] Technical Reference Manual there is a layer of digital logic in front of the QSPI/OSPI Drive when used in DAC mode. This is part of the Flash Subsystem (FSS) which provides access to external Flash devices. The FSS0_0_SYSCONFIG Register (Offset = 4h) has a BIT Field for OSPI_32B_DISABLE_MODE which has a Reset value = 0. This means, OSPI 32bit mode enabled by default. Thus, by default controller operates in 32 bit mode causing it to always align all data to 4 bytes from a 4byte aligned address. In some flash chips like cypress for example if we try to read some regs in DAC mode then it keeps sending the value of the first register that was requested and inorder to read the next reg, we have to stop and re-initiate a new transaction. This causes wrong register values to be read than what is desired when registers are read in DAC mode. Hence if the data.nbytes is very less then prefer STIG mode for such small reads. [0] https://www.ti.com/lit/ug/spruiv7a/spruiv7a.pdf Signed-off-by: Dhruva Gole <d-gole@ti.com> --- drivers/spi/spi-cadence-quadspi.c | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) -- 2.25.1