66AK2H06: IPC interrupt delay (Again)

Part Number: 66AK2H06

Hi xeprts,

This is a follow-up to the previous thread.

Initially, the issue was observed when an IPC interrupt was generated to the same DSP core (Core0 → Core0).However, we have since found that the same issue can also occur when the IPC interrupt is generated to a different DSP core (Core0 → Core1).

Q1:Is it possible that a delay of about 500 usec occurs between the IPC interrupt request and the start of IPC interrupt processing on different core due to the specifications of the 66AK2H06?

Q2. If a delay of about 500 usec does not normally occur in [Q1], could you please tell me the possible causes and the points I should check? 

Best regards,
O.H

  • Hello,
    Thank you for your inquiry. **x1125363@ti.com** is currently on leave and will be available on **2026-08-28**.
    Your query will be addressed upon their return. We appreciate your patience and understanding.
    Best regards,
    TI E2E Support Team
    ---
    *This is an automated notification.*

  • Hi,

    The assigned expert is currently on vacation. Kindly expect a response early next week.

    Thank you for your patience.

    Best Regards, 

    Sudheer

  • Hi O.H,

    Apologies for the delay. I was on Leave.

    To narrow down the cause, could you please let us know how the ~500 µs is being measured? Specifically:

    Which timestamp source are you using & Where are the two measurement points taken? If possible, a small snippet of your test code for the send and receive paths (or the measurement portion only), and/or a screenshot, would be very helpful.

    Regards,

    Betsy Varughese

  • Hi Betsy Varughese,

    I apologize. I accidentally deleted the previous information. Even though I shouldn't have had the permissions to do so, your post ended up getting deleted as well... I will send the information again, including the latest updates.

    =============

    We would like to provide additional findings regarding the IPC interrupt delay issue discussed in this thread.

    Occurrence rate

    Approximately 200 boards have been produced so far, and the issue has been observed on 3 boards.
    For the board currently available to us, all of the following cases have been reproduced:

    • Core0 → Core0
    • Core1 → Core1
    • Core0 → Core1
    • Core1 → Core0

    We believe that the other two affected boards also exhibited both the same-core and cross-core IPC delay issues.

    Additional test results

    Initially, we observed the issue when an IPC interrupt was generated to the same DSP core (for example, Core0 → Core0).
    However, further testing has confirmed that the same behavior also occurs between different DSP cores (Core0 → Core1 and Core1 → Core0).
    Furthermore, we confirmed that the issue also occurs for Core1 → Core1 IPC interrupts.

    Investigation of IFR status

    We investigated the Interrupt Flag Register (IFR) during the occurrence of the issue.
    Our observation suggests that the delay occurs before the corresponding IFR bit is set on the receiving core.
    In other words, after writing to IPCGR and issuing the IPC interrupt request, the IFR bit on the destination core is not asserted immediately and is delayed by approximately 500 µs when the issue occurs.

    Comparison with direct interrupt generation

    As an additional experiment, we generated an interrupt by directly writing to the interrupt mechanism instead of using the IPC interrupt path.
    In this case, the approximately 500 µs delay was not observed.
    Therefore, the issue appears to be specific to the IPC interrupt path.

    Any guidance on additional registers, status information, or debug points that should be checked would be greatly appreciated.

    ==================

    Best regards,
    O.H

    #include "inspdef.h"
    
    extern cregister volatile unsigned int IER;
    extern cregister volatile unsigned int CSR;
    extern cregister volatile unsigned int ISTP;
    extern cregister volatile unsigned int IFR;
    extern cregister volatile unsigned int ICR;
    extern cregister volatile unsigned int TSCL;
    extern cregister volatile unsigned int TSCH;
    extern cregister volatile unsigned int DNUM;
    
    extern cregister volatile unsigned int ITSR;
    extern cregister volatile unsigned int IRP;
    extern cregister volatile unsigned int ISR;
    
    volatile unsigned int	start_flag = 0;
    volatile int			start;
    volatile int			end;
    volatile int			diff;
    volatile int			max = 0x80000000;
    volatile int			min = 0x7FFFFFFF;
    volatile int			ipc_cnt_old;
    
    //------------------------------------------------------------------------------
    // The latter half of core0INRAM is used for the handshake
    //------------------------------------------------------------------------------
    #define	IPC_CNT		(*(volatile int *)0x10880000)
    #define	IPC_FLAG	(*(volatile int *)0x10880004)
    
    //------------------------------------------------------------------------------
    void main( void )
    {
    	systemInit();
    
    	IPC_CNT		= 0;
    
    	while( 1 ){
    		if( start_flag == 1 ){
    			break;
    		}
    	}
    
    	//------------------------------------------------------------------------------
    	while( 1 ){
    
    		ipc_cnt_old	= IPC_CNT;
    
    		IPC_FLAG	= 0;
    
    		//------------------------------------------------------------------------------
    		// Clock when an interrupt request is made to core1
    		//------------------------------------------------------------------------------
    		start		= TSCL;
    
    		//------------------------------------------------------------------------------
    		// IPC interrupt request made to core1
    		//------------------------------------------------------------------------------
    		ipcRequest( 1, 1 );
    
    		while( 1 ){
    			if( IPC_FLAG == 1 ){
    				//------------------------------------------------------------------------------
    				// IPC interrupt detected by core1
    				//------------------------------------------------------------------------------
    				break;
    			}
    		}
    
    		//------------------------------------------------------------------------------
    		// Clock when interrupt is detected by core1
    		//------------------------------------------------------------------------------
    		end			= TSCL;
    
    		//------------------------------------------------------------------------------
    		// Comparison with the second and subsequent IPC interrupts
    		//------------------------------------------------------------------------------
    		if( IPC_CNT != 0 ){
    			diff = (int)( end - start );
    			if( max <= diff ){
    				max = diff;
    			}
    			if( min >= diff ){
    				min = diff;
    			}
    		}
    
    		while( 1 ){
    			if( IPC_CNT == ( ipc_cnt_old + 1 )){
    				//------------------------------------------------------------------------------
    				// Completion of IPC interrupt processing by core1
    				//------------------------------------------------------------------------------
    				break;
    			}
    		}
    
    		waitUsec( 2000 );
    
    		if( start_flag != 1 ){
    			//------------------------------------------------------------------------------
    			// Termination notification detected
    			//------------------------------------------------------------------------------
    			break;
    		}
    	}
    	//------------------------------------------------------------------------------
    
    	//------------------------------------------------------------------------------
    	// End
    	//------------------------------------------------------------------------------
    	start_flag = 2;
    }
    //------------------------------------------------------------------------------
    
     
    #include "inspdef.h"
    
    extern cregister volatile unsigned int IER;
    extern cregister volatile unsigned int CSR;
    extern cregister volatile unsigned int ISTP;
    extern cregister volatile unsigned int IFR;
    extern cregister volatile unsigned int ICR;
    extern cregister volatile unsigned int TSCL;
    extern cregister volatile unsigned int TSCH;
    extern cregister volatile unsigned int DNUM;
    
    extern cregister volatile unsigned int ITSR;
    extern cregister volatile unsigned int IRP;
    extern cregister volatile unsigned int ISR;
    
    volatile unsigned int	start_flag = 0;
    
    //------------------------------------------------------------------------------
    // The latter half of core0INRAM is used for the handshake.
    //------------------------------------------------------------------------------
    #define	IPC_CNT		(*(volatile int *)0x10880000)
    #define	IPC_FLAG	(*(volatile int *)0x10880004)
    
    //------------------------------------------------------------------------------
    interrupt void ipc_int( void )
    {
    	IPC_FLAG = 1;		// Interrupt occurrence is notified to core0
    
    	waitUsec( 1500 );	// 1.5ms
    
    	ipcClear( 1 );		// IPC clear
    
    	IPC_CNT++;
    }
    //------------------------------------------------------------------------------
    void main( void )
    {
    	systemInit();
    
    	while( 1 ){
    		if( start_flag == 1 ){
    			break;
    		}
    	}
    
    	ipcSetInter( 6, ipc_int );			// IPC interrupt settings are configured.
    
    	enableGInter();	//GIE enable	CSR |= 1;
    
    	//------------------------------------------------------------------------------
    	while( 1 ){
    		if( start_flag != 1 ){
    			//------------------------------------------------------------------------------
    			// Termination notification is detected.
    			//------------------------------------------------------------------------------
    			break;
    		}
    	}
    	//------------------------------------------------------------------------------
    
    	disableGInter();	// GIE disable
    	disableLinter( 6 );	// IER disable
    	ipcClear( 1 );		// IPC clear
    
    	//------------------------------------------------------------------------------
    	// End
    	//------------------------------------------------------------------------------
    	start_flag = 2;
    }
    //------------------------------------------------------------------------------
    
     
    #define CORE_TOP_ADR	0x02620000
    
    #define CORE_IPCGR_CORE(core)	(*(volatile unsigned int *)(CORE_TOP_ADR + 0x240 + 0x4 * (core)))
    
    /*====================================================================
    	Inter-DSP interrupt Request	Interrupt sender
    ====================================================================*/
    void ipcRequest(int coreNum, int srcId)
    {
    	CORE_IPCGR_CORE(coreNum) = 1 | (1 << (4 + srcId));
    }
    

  • Hi Betsy Varughese,

    To narrow down the cause, could you please let us know how the ~500 µs is being measured? Specifically:

    They are using the DSP internal timer (TSCL) as the timestamp source for our measurements.

    In the test loop, there is a 2 ms delay between issuing the IPC interrupt request and starting the next IPC interrupt detection cycle. Therefore, the request rate is approximately 500 Hz.

    Based on our observations so far, the delay varies between approximately 500 µs and 700 µs.

    They performed the additional debugging procedure that you requested (running on DSP Core0):

    1. Timestamp measurement #1 (read TSCL)
    2. IPCGR write (IPC interrupt request to DSP Core1)
    3. IPCGR read-back
    4. Timestamp measurement #2 (read TSCL)

    The accesses in steps [2] and [3] were both to IPCGR1.
    The value written in step [2] was 0x00000021.
    For the final cycle of the test loop, the value read from IPCGR1 in step [3], as well as in subsequent reads after step [4], was 0x00000020.
    The interval between timestamp measurements #1 and #2 ranged from:

    • Maximum: 633.533 µs (TSCL difference: 760240)
    • Minimum: 96 ns (TSCL difference: 115)

    Best regards,
    O.H

  • Hi O.H,

    No worries.

    They are using the DSP internal timer (TSCL) as the timestamp source for our measurements.

    Understood. Thank you for the detailed measurements, that is very helpful.
    1. Could you add a third timestamp between the write and the read-back, so the two are timed separately. The write does not stall the core but the read does, so we would like to know which of the two is waiting.
    2. Could you tell us the version of the board — the hardware revision of the affected units, and whether the three failing boards are the same revision as the ones that work.
    Regards,
    Betsy Varughese
  • Hi Betsy Varughese,

    Thank you for your supprots.

    Could you add a third timestamp between the write and the read-back, so the two are timed separately. The write does not stall the core but the read does, so we would like to know which of the two is waiting.

    They performed the timing measurements using the following sequence:

    1. Timestamp measurement #1
    2. IPCGR write (IPC interrupt request)
    3. Timestamp measurement #2
    4. IPCGR read-back
    5. Timestamp measurement #3

    The measurement results were as follows:

    • Maximum interval between [1] and approximately 3 ns
    • Maximum interval between [1] and approximately 650 µs

    Based on these results, the IPCGR write operation appears to complete immediately, while the read-back operation tends to take a significant amount of time.

    Could you tell us the version of the board — the hardware revision of the affected units, and whether the three failing boards are the same revision as the ones that work.

    Regarding the board version, our understanding is that the affected units and the units operating normally are the same hardware revision.
    Unfortunately, it is currently difficult for us to determine the silicon (chip) revision of the affected devices.

    Best regards,
    O.H

  • Hi O.H,

    Thanks for sharing the requested info.

    Is this a custom board or a TI EVM? Can you run the same on a TI K2H EVM (EVMK2H) and tell us whether the delay reproduces? & please share a register dump from one failing board and one good board, for comparison. 

    Could you confirm that waitUsec() reads only TSCL/TSCH, and does not access any other register?

    Regards,

    Betsy Varughese

  • Hi Betsy Varughese,

    Soory for late reply. We have asked the customer to verify the details, so please bear with us a little longer.

    Best regards,
    O.H

  • Hi O.H,

    Sure, I will wait for the response. In the meantime, I’ll move this thread to the “Waiting for Customer” stage.
    Regards,
    Betsy Varughese
  • Hi 

    Customer feedback below.

    Is this a custom board or a TI EVM? Can you run the same on a TI K2H EVM (EVMK2H) and tell us whether the delay reproduces? & please share a register dump from one failing board and one good board, for comparison. 

    We were unable to reproduce this issue on a TI K2H EVM (EVMK2H).
    Q3: Could you please advise which registers you would like us to include in the register dump?

    Could you confirm that waitUsec() reads only TSCL/TSCH, and does not access any other register?

    Regarding the timer implementation, only TSCL and TSCH registers are used.

    As additional information, we observed that the occurrence of the delay appears to be related to whether the USB controller is enabled in the Linux device tree running on the ARM cores.

    Our observations are as follows:

    • USB controller enabled in the Linux device tree → No delay observed
    • USB controller disabled in the Linux device tree → Delay observed

    Q4: Could you please advise how the presence or absence of the USB controller in the Linux device tree might affect the IPC mechanism?
    In addition, what factors could explain why this issue occurs only on certain devices, while other devices operate normally under the same hardware revision and software configuration?

    Any guidance would be greatly appreciated.

    Best regards,
    O.H

  • Hi O.H,

    Thanks for the details. Let me review the information provided, and I will update you shortly.
    Regards,
    Betsy Varughese
  • Hi Betsy Varughese,

    Our observations are as follows:

    • USB controller enabled in the Linux device tree → No delay observed
    • USB controller disabled in the Linux device tree → Delay observed

    We apologize, but there was an error in our previous.

    Correct:

    • USB controller enabled in the Linux device tree → Delay observed
    • USB controller disabled in the Linux device tree → No delay observed
    Regarding the EVMK2H, they only performed the evaluation in a non-Linux environment.

    One notable difference between the EVMK2H and their product is the USB-related circuitry. On their board, the SoC USB-related pins (VPH, VDDUSB, VP, and VPTX) are tied to GND, whereas the EVMK2H is configured differently. They have observed the following correlation:

    • USB controller enabled in the Linux device tree:
      • IPC delay observed
      • "hub 1-0:1.0: connect-debounce failed, port 1 disabled" message observed
    • USB controller disabled in the Linux device tree:
      • IPC delay not observed
      • Above USB log message not observed

    Based on these observations, they suspect that the USB subsystem may be related to the IPC delay. In addition, they have not yet been able to determine why the IPC delay occurs on some SoCs while other SoCs operate normally under seemingly identical conditions. They are also concerned about whether devices that are currently operating without any IPC delay issues could potentially develop the same symptom over time due to aging or long-term reliability effects.

    Could TI comment on whether an abnormal USB PHY condition, USB controller initialization failure, or related hardware configuration might have any impact on IPC event delivery or IPCGR access timing?

    Best regards,
    O.H

  • Hi O.H,

    Since the USB supply pins are grounded, could you please confirm whether USB is intended to be used on this product? If USB is not required, it is recommended to disable the USB controller in the device tree.

    To help us investigate the delay, could you please confirm the following:

    1. Please share the complete boot log from an affected board with USB enabled.
    2. Are the affected and unaffected boards using the same device tree?
    3. With USB disabled, does the delay remain absent during a long-duration test?

    If USB is required, please let us know and we can provide the required hardware configuration details.

    You can also refer the section 2.2.8 USB Power Supplies  - https://www.ti.com/lit/an/sprabv0/sprabv0.pdf?

    Best regards,

    Betsy Varughese

  • Hi Betsy Varughese,

    Customer feedback below.

    Since the USB supply pins are grounded, could you please confirm whether USB is intended to be used on this product? If USB is not required, it is recommended to disable the USB controller in the device tree.

    I do not plan to use the USB function.

    Please share the complete boot log from an affected board with USB enabled.

    I have attached the boot log (startup_log.txt). Around line 308, it contains the message: "hub 1-0:1.0: connect-debounce failed, port 1 disabled".

    K2HK_EMAC, K2HK_EMAC1, K2HK_EMAC2, K2HK_EMAC3
    Hit any key to stop autoboot:  0
    Creating 1 MTD partitions on "nand0":
    0x000000180000-0x000020000000 : "mtd=2"
    UBI: attaching mtd1 to ubi0
    UBI: physical eraseblock size:   131072 bytes (128 KiB)
    UBI: logical eraseblock size:    126976 bytes
    UBI: smallest flash I/O unit:    2048
    UBI: VID header offset:          2048 (aligned 2048)
    UBI: data offset:                4096
    UBI: attached mtd1 to ubi0
    UBI: MTD device name:            "mtd=2"
    UBI: MTD device size:            510 MiB
    UBI: number of good PEBs:        4080
    UBI: number of bad PEBs:         4
    UBI: max. allowed volumes:       128
    UBI: wear-leveling threshold:    4096
    UBI: number of internal volumes: 1
    UBI: number of user volumes:     2
    UBI: available PEBs:             0
    UBI: total number of reserved PEBs: 4080
    UBI: number of PEBs reserved for bad PEB handling: 40
    UBI: max/mean erase counter: 3/1
    UBIFS: mounted UBI device 0, volume 0, name "boot"
    UBIFS: mounted read-only
    UBIFS: file system size:   4698112 bytes (4588 KiB, 4 MiB, 37 LEBs)
    UBIFS: journal size:       1142785 bytes (1116 KiB, 1 MiB, 8 LEBs)
    UBIFS: media format:       w4/r0 (latest is w4/r0)
    UBIFS: default compressor: LZO
    UBIFS: reserved for root:  0 bytes (0 KiB)
    Loading file 'uImage-k2hk-evm.dtb' to addr 0x87000000 with size 53609 (0x0000d169)...
    Done
    Loading file 'skern-keystone-evm.bin' to addr 0x0c5f0000 with size 45056 (0x0000b000)...
    Done
    Loading file 'uImage-keystone-evm.bin' to addr 0x88000000 with size 4146736 (0x003f4630)...
    Done
    K2_BM_15. 07-nogit SoC:k2hk built:20:49:49, Jan  6 2020
    
    ## installed monitor, freq [200000000], status 0
    ## Booting kernel from Legacy Image at 88000000 ...
       Image Name:   Linux-3.10.72
       Created:      2023-09-29   1:56:35 UTC
       Image Type:   ARM Linux Kernel Image (uncompressed)
       Data Size:    4146672 Bytes = 4 MiB
       Load Address: 80008000
       Entry Point:  80008000
       Verifying Checksum ... OK
    ## Flattened Device Tree blob at 87000000
       Booting using the fdt blob at 0x87000000
       Loading Kernel Image ... OK
    OK
       Using Device Tree in place at 87000000, end 87010168
    
    Starting kernel ...
    
    [    0.000000] Booting Linux on physical CPU 0x0
    [    0.000000] Linux version 3.10.72 (root@ubuntu) (gcc version 4.7.3 20130226 (prerelease) (crosstool-NG linaro-1.13.1-4.7-2013.03-20130313 - Linaro GCC 2013.03) ) #8 Ver.1.0.0 SMP Thu Sep 28 18:56:28 PDT 2023
    [    0.000000] CPU: ARMv7 Processor [412fc0f4] revision 4 (ARMv7), cr=30c7387d
    [    0.000000] CPU: PIPT / VIPT nonaliasing data cache, PIPT instruction cache
    [    0.000000] Machine: KeyStone2, model: Texas Instruments Keystone 2 SoC (DTS Ver.1.0.0)
    [    0.000000] switching to high address space at 0x800000000
    [    0.000000] cma: CMA: reserved 16 MiB at 0e000000
    [    0.000000] Memory policy: ECC disabled, Data cache writealloc
    [    0.000000] PERCPU: Embedded 8 pages/cpu @c145d000 s11264 r8192 d13312 u32768
    [    0.000000] Built 1 zonelists in Zone order, mobility grouping on.  Total pages: 392736
    [    0.000000] Kernel command line: console=ttyS0,115200n8 vmalloc=768M rootwait=1 rootfstype=ubifs root=ubi0:rootfs ro ubi.mtd=2,2048
    [    0.000000] PID hash table entries: 1024 (order: 0, 4096 bytes)
    [    0.000000] Dentry cache hash table entries: 32768 (order: 5, 131072 bytes)
    [    0.000000] Inode-cache hash table entries: 16384 (order: 4, 65536 bytes)
    [    0.000000] Memory: 1536MB = 1536MB total
    [    0.000000] Memory: 1535252k/1535252k available, 37612k reserved, 1327104K highmem
    [    0.000000] Virtual kernel memory layout:
    [    0.000000]     vector  : 0xffff0000 - 0xffff1000   (   4 kB)
    [    0.000000]     fixmap  : 0xfff00000 - 0xfffe0000   ( 896 kB)
    [    0.000000]     vmalloc : 0xcf800000 - 0xff000000   ( 760 MB)
    [    0.000000]     lowmem  : 0xc0000000 - 0xcf000000   ( 240 MB)
    [    0.000000]     pkmap   : 0xbfe00000 - 0xc0000000   (   2 MB)
    [    0.000000]     modules : 0xbf000000 - 0xbfe00000   (  14 MB)
    [    0.000000]       .text : 0xc0008000 - 0xc0751574   (7462 kB)
    [    0.000000]       .init : 0xc0752000 - 0xc07a4c00   ( 331 kB)
    [    0.000000]       .data : 0xc07a6000 - 0xc08092d8   ( 397 kB)
    [    0.000000]        .bss : 0xc08092d8 - 0xc083afbc   ( 200 kB)
    [    0.000000] SLUB: HWalign=64, Order=0-3, MinObjects=0, CPUs=2, Nodes=1
    [    0.000000] Hierarchical RCU implementation.
    [    0.000000] NR_IRQS:16 nr_irqs:16 16
    [    0.000000] ipc irq: irqchip registered, range 512-539
    [    0.000000] Main PLL clk (1200000000 Hz), parent (100000000 Hz),postdiv = 2, mult = 23, prediv = 0
    [    0.000000] Generic PLL clk (1200000000 Hz), parent (100000000 Hz),postdiv = 2, mult = 23, prediv = 0
    [    0.000000] Generic PLL clk (983333294 Hz), parent (100000000 Hz),postdiv = 2, mult = 235, prediv = 11
    [    0.000000] Generic PLL clk (400000000 Hz), parent (100000000 Hz),postdiv = 4, mult = 15, prediv = 0
    [    0.000000] Generic PLL clk (400000000 Hz), parent (100000000 Hz),postdiv = 4, mult = 15, prediv = 0
    [    0.000000] Architected local timer running at 200.00MHz (phys).
    [    0.000000] Switching to timer-based delay loop
    [    0.000000] sched_clock: ARM arch timer >56 bits at 200000kHz, resolution 640/128ns
    [    0.000000] keystone timer clock @200000000 MHz
    [    0.000000] sched_clock: 32 bits at 100 Hz, resolution 10000000ns, wraps every 4294967286ms
    [    0.000000] Console: colour dummy device 80x30
    [   15.727574] Calibrating delay loop (skipped), value calculated using timer frequency.. 400.00 BogoMIPS (lpj=2000000)
    [   15.727583] pid_max: default: 4096 minimum: 301
    [   15.727714] Mount-cache hash table entries: 512
    [   15.736590] CPU: Testing write buffer coherency: ok
    [   15.736763] CPU0: thread -1, cpu 0, socket 0, mpidr 80000000
    [   15.736773] psci: probing function IDs from device-tree
    [   15.736785] Setting up static identity map for 0xc0528520 - 0xc0528578
    [   15.742090] CPU1: Booted secondary processor
    [   15.742116] CPU1: thread -1, cpu 1, socket 0, mpidr 80000001
    [   15.742188] Brought up 2 CPUs
    [   15.742200] SMP: Total of 2 processors activated (800.00 BogoMIPS).
    [   15.742205] CPU: All CPU(s) started in SVC mode.
    [   15.742621] devtmpfs: initialized
    [   15.755026] NET: Registered protocol family 16
    [   15.755851] DMA: preallocated 256 KiB pool for atomic coherent allocations
    [   15.768304] No ATAGs?
    [   15.768319] hw-breakpoint: found 5 (+1 reserved) breakpoint and 4 watchpoint registers.
    [   15.768325] hw-breakpoint: maximum watchpoint size is 8 bytes.
    [   15.778487] bio: create slab <bio-0> at 0
    [   15.778860] keystone-pcie: keystone_pcie_rc_init - start
    [   15.778882] keystone2_pcie_serdes_setup for domain 0
    [   15.781004] keystone2_pcie_serdes_setup done domain 0, en_link_train = 1
    [   15.781033] keystone-pcie: MEM 0x0000000080000000..0x000000008fffffff -> 0x0000000080000000
    [   15.781043] keystone-pcie: IO 0x0000000023250000..0x0000000023253fff -> 0x0000000000000000
    [   15.781074] keystone-pcie: pcie - number of legacy irqs = 4
    [   15.781118] keystone-pcie: pcie - number of MSI host irqs = 8, msi_irqs = 32
    [   15.887442] keystone-pcie: Doing PCI Setup...Done
    [   15.887447] keystone-pcie: Starting PCI scan, nr 0...
    [   15.887576] PCI host bridge to bus 0000:00
    [   15.887587] pci_bus 0000:00: root bus resource [mem 0x80000000-0x8fffffff]
    [   15.887594] pci_bus 0000:00: root bus resource [io  0x0000-0x3fff]
    [   15.887601] pci_bus 0000:00: No busn resource found for root bus, will use [bus 00-ff]
    [   15.887650] PCI: bus0: Fast back to back transfers enabled
    [   15.887670] keystone-pcie: Ending PCI scan...
    [   15.887677] keystone-pcie: keystone_pcie_rc_init - end
    [   15.887937] vgaarb: loaded
    [   15.888232] SCSI subsystem initialized
    [   15.888806] usbcore: registered new interface driver usbfs
    [   15.888884] usbcore: registered new interface driver hub
    [   15.888978] usbcore: registered new device driver usb
    [   15.890055] pps_core: LinuxPPS API ver. 1 registered
    [   15.890060] pps_core: Software ver. 5.3.6 - Copyright 2005-2007 Rodolfo Giometti <giometti@linux.it>
    [   15.890121] PTP clock support registered
    [   15.890262] keystone-hwqueue hwqueue.36: qmgr start queue 0, number of queues 8192
    [   15.890338] keystone-hwqueue hwqueue.36: added qmgr start queue 0, num of queues 8192, reg_peek cf840000, reg_status cf806000, reg_config cf808000, reg_region cf80a000, reg_push cf880000, reg_pop cf8c0000
    [   15.890347] keystone-hwqueue hwqueue.36: qmgr start queue 8192, number of queues 8192
    [   15.890417] keystone-hwqueue hwqueue.36: added qmgr start queue 8192, num of queues 8192, reg_peek cf900000, reg_status cf80c400, reg_config cf80e000, reg_region cf81a000, reg_push cf940000, reg_pop cf980000
    [   15.891184] keystone-hwqueue hwqueue.36: qos: sched port @8096, drop sched @8000
    [   15.892207] keystone-hwqueue hwqueue.36: qos: sched port @6496, drop sched @6400
    [   15.893160] keystone-hwqueue hwqueue.36: added pool pool-rio: 1024 descriptors of size 128
    [   15.893169] keystone-hwqueue hwqueue.36: added pool pool-xge: 1024 descriptors of size 128
    [   15.893178] keystone-hwqueue hwqueue.36: added pool pool-crypto: 1024 descriptors of size 128
    [   15.893187] keystone-hwqueue hwqueue.36: added pool pool-net: 11264 descriptors of size 128
    [   15.893195] keystone-hwqueue hwqueue.36: added pool pool-udma: 1920 descriptors of size 256
    [   15.895276] keystone-hwqueue hwqueue.36: registered queues 0-16383
    [   15.895550] keystone-hwqueue hwqueue.36: qos version 0x200010b, magic valid
    [   15.896038] keystone-hwqueue hwqueue.36: qos version 0x200010b, magic valid
    [   15.903875] keystone-pktdma 2a08000.pktdma: registered 32 logical channels, flows 32, tx chans: 32, rx chans: 32, loopback
    [   15.907395] keystone-pktdma 2004000.pktdma: registered 33 logical channels, flows 32, tx chans: 9, rx chans: 24
    [   15.907919] keystone-pktdma 2fa1000.pktdma: registered 4 logical channels, flows 32, tx chans: 16, rx chans: 16
    [   15.908044] Switching to clocksource arch_sys_counter
    [   15.925909] NET: Registered protocol family 2
    [   15.926367] TCP established hash table entries: 2048 (order: 2, 16384 bytes)
    [   15.926401] TCP bind hash table entries: 2048 (order: 2, 16384 bytes)
    [   15.926433] TCP: Hash tables configured (established 2048 bind 2048)
    [   15.926474] TCP: reno registered
    [   15.926483] UDP hash table entries: 128 (order: 0, 4096 bytes)
    [   15.926493] UDP-Lite hash table entries: 128 (order: 0, 4096 bytes)
    [   15.926649] NET: Registered protocol family 1
    [   15.926796] RPC: Registered named UNIX socket transport module.
    [   15.926802] RPC: Registered udp transport module.
    [   15.926806] RPC: Registered tcp transport module.
    [   15.926811] RPC: Registered tcp NFSv4.1 backchannel transport module.
    [   15.927205] hw perfevents: enabled with ARMv7 Cortex-A15 PMU driver, 7 counters available
    [   16.004455] bounce pool size: 64 pages
    [   16.011497] Installing knfsd (copyright (C) 1996 okir@monad.swb.de).
    [   16.011653] NTFS driver 2.1.30 [Flags: R/O].
    [   16.011869] jffs2: version 2.2. (NAND) c 2001-2006 Red Hat, Inc.
    [   16.012145] msgmni has been set to 438
    [   16.013020] NET: Registered protocol family 38
    [   16.013199] Block layer SCSI generic (bsg) driver version 0.4 loaded (major 252)
    [   16.013206] io scheduler noop registered
    [   16.013211] io scheduler deadline registered
    [   16.013325] io scheduler cfq registered (default)
    [   16.014755] keystone-udma udma0.37: registered udma device udma0
    [   16.069898] Serial: 8250/16550 driver, 4 ports, IRQ sharing disabled
    [   16.071383] 2530c00.serial: ttyS0 at MMIO 0x2530c00 (irq = 309) is a 16550A
    [   16.898924] console [ttyS0] enabled
    [   16.903034] 2531000.serial: ttyS1 at MMIO 0x2531000 (irq = 312) is a 16550A
    [   16.912953] loop: module loaded
    [   16.916231] at24 0-0050: 131072 byte 24c1024 EEPROM, writable, 1 bytes/write
    [   16.924581] Generic platform RAM MTD, (c) 2004 Simtec Electronics
    [   16.931599] ONFI param page 0 valid
    [   16.935097] ONFI flash detected
    [   16.938266] NAND device: Manufacturer ID: 0x01, Chip ID: 0xac (AMD/Spansion S34MS04G2), 512MiB, page size: 2048, OOB size: 128
    [   16.949996] Bad block table found at page 262080, version 0x01
    [   16.956360] Bad block table found at page 262016, version 0x01
    [   16.962533] 3 ofpart partitions found on MTD device 30000000.nand
    [   16.968652] Creating 3 MTD partitions on "30000000.nand":
    [   16.974069] 0x000000000000-0x000000100000 : "u-boot"
    [   16.979670] 0x000000100000-0x000000180000 : "params"
    [   16.985228] 0x000000180000-0x000020000000 : "ubifs"
    [   16.990884] davinci_nand 30000000.nand: controller rev. 2.5
    [   16.997057] spi_davinci 21000400.spi: master is unqueued, this is deprecated
    [   17.004142] spi_davinci 21000400.spi: Controller at 0xcf87a400
    [   17.010290] spi_davinci 21000600.spi: master is unqueued, this is deprecated
    [   17.017364] spi_davinci 21000600.spi: Controller at 0xcf87c600
    [   17.023519] spi_davinci 21000800.spi: master is unqueued, this is deprecated
    [   17.030605] spi_davinci 21000800.spi: Controller at 0xcf87e800
    [   17.037540] tun: Universal TUN/TAP device driver, 1.6
    [   17.042618] tun: (C) 1999-2004 Max Krasnyansky <maxk@qualcomm.com>
    [   17.098078] davinci_mdio 2090300.mdio: davinci mdio revision 1.5
    [   17.104103] libphy: 2090300.mdio: probed
    [   17.108605] mdio_bus 2090300.mdio: cannot get PHY at address 1
    [   17.114459] davinci_mdio 2090300.mdio: phy[0]: device 2090300.mdio:00, driver Marvell 88E1510
    [   17.123316] keystone-netcp 2090000.netcp: number of serdeses 1
    [   17.129192] keystone-netcp 2090000.netcp: Missing cpts_clock_mult property in the DT.
    [   17.137049] keystone-netcp 2090000.netcp: Missing cpts_clock_shift property in the DT.
    [   17.145000] keystone-netcp 2090000.netcp: Missing cpts_clock_div property in the DT.
    [   17.153716] keystone-netcp 2090000.netcp: cpsw serdes 0 config done lanes(mask) 0x1
    [   17.162232] keystone-netcp 2090000.netcp: Created interface "eth0"
    [   17.168451] keystone-netcp 2090000.netcp: dma_chan_name nettx0
    [   17.175069] keystone-netcp 2090000.netcp: Created interface "eth1"
    [   17.181289] keystone-netcp 2090000.netcp: dma_chan_name nettx1
    [   17.187149] keystone-netcp 2090000.netcp: could not find interface-2
    [   17.193538] keystone-netcp 2090000.netcp: could not find interface-3
    [   17.200042] keystone-netcp 2090000.netcp: pdsp 0 firmware: keystone/pa_pdsp0_classify1.fw
    [   17.208257] keystone-netcp 2090000.netcp: pdsp 1 firmware: keystone/pa_pdsp1_classify1.fw
    [   17.216464] keystone-netcp 2090000.netcp: pdsp 2 firmware: keystone/pa_pdsp2_classify1.fw
    [   17.224676] keystone-netcp 2090000.netcp: pdsp 3 firmware: keystone/pa_pdsp3_classify2.fw
    [   17.232890] keystone-netcp 2090000.netcp: pdsp 4 firmware: keystone/pa_pdsp45_pam.fw
    [   17.240667] keystone-netcp 2090000.netcp: pdsp 5 firmware: keystone/pa_pdsp45_pam.fw
    [   17.283414] keystone-netcp 2090000.netcp: Using Packet Accelerator Firmware version 0x03000105
    [   17.292068] keystone-netcp 2090000.netcp: pa_clk_rate(163888882 HZ),mult(24993),shift(12)
    [   17.303135] dwc3 2690000.dwc3: no usb2 phy configured
    [   17.308279] platform 2690000.dwc3: Driver dwc3 requests probe deferral
    [   17.315262] usbcore: registered new interface driver usb-storage
    [   17.321721] mousedev: PS/2 mouse device common for all mice
    [   17.327502] i2c /dev entries driver
    [   17.331659] watchdog 22f0080.wdt: heartbeat 60 sec
    [   17.347854] keystone-crypto 20c0000.crypto: crypto accelerator enabled
    [   17.354738] usbcore: registered new interface driver usbhid
    [   17.360336] usbhid: USB HID core driver
    [   17.364430] rproc-user 2620040.dsp0: registered misc device dsp0
    [   17.370623] rproc-user 2620044.dsp1: registered misc device dsp1
    [   17.376801] rproc-user 2620048.dsp2: registered misc device dsp2
    [   17.382986] rproc-user 262004c.dsp3: registered misc device dsp3
    [   17.389169] rproc-user 2620050.dsp4: registered misc device dsp4
    [   17.395344] rproc-user 2620054.dsp5: registered misc device dsp5
    [   17.401525] rproc-user 2620058.dsp6: registered misc device dsp6
    [   17.407698] rproc-user 262005c.dsp7: registered misc device dsp7
    [   17.413754] rproc-user dspmem.38: kick gpio
    [   17.418076] rproc-user dspmem.38: registered misc device dspmem
    [   17.424319] oprofile: using arm/armv7-ca15
    [   17.428545] ipip: IPv4 over IPv4 tunneling driver
    [   17.433643] gre: GRE over IPv4 demultiplexor driver
    [   17.438546] ip_gre: GRE over IPv4 tunneling driver
    [   17.444055] TCP: cubic registered
    [   17.447378] Initializing XFRM netlink socket
    [   17.452075] NET: Registered protocol family 10
    [   17.457461] NET: Registered protocol family 17
    [   17.461941] NET: Registered protocol family 15
    [   17.466448] Bridge firewalling registered
    [   17.470480] 8021q: 802.1Q VLAN Support v1.8
    [   17.476406] sctp: Hash tables configured (established 65536 bind 65536)
    [   17.483144] NET: Registered protocol family 41
    [   17.487723] VFP support v0.3: implementor 41 architecture 4 part 30 variant f rev 0
    [   17.495432] Registering SWP/SWPB emulation handler
    [   17.702029] xhci-hcd xhci-hcd.0.auto: xHCI Host Controller
    [   17.707543] xhci-hcd xhci-hcd.0.auto: new USB bus registered, assigned bus number 1
    [   17.715639] xhci-hcd xhci-hcd.0.auto: irq 425, io mem 0x02690000
    [   17.721741] usb usb1: New USB device found, idVendor=1d6b, idProduct=0002
    [   17.728562] usb usb1: New USB device strings: Mfr=3, Product=2, SerialNumber=1
    [   17.735808] usb usb1: Product: xHCI Host Controller
    [   17.740706] usb usb1: Manufacturer: Linux 3.10.72 xhci-hcd
    [   17.746207] usb usb1: SerialNumber: xhci-hcd.0.auto
    [   17.751552] hub 1-0:1.0: USB hub found
    [   17.755318] hub 1-0:1.0: 1 port detected
    [   17.759441] xhci-hcd xhci-hcd.0.auto: xHCI Host Controller
    [   17.764949] xhci-hcd xhci-hcd.0.auto: new USB bus registered, assigned bus number 2
    [   17.772729] usb usb2: New USB device found, idVendor=1d6b, idProduct=0003
    [   17.779548] usb usb2: New USB device strings: Mfr=3, Product=2, SerialNumber=1
    [   17.786794] usb usb2: Product: xHCI Host Controller
    [   17.791692] usb usb2: Manufacturer: Linux 3.10.72 xhci-hcd
    [   17.797195] usb usb2: SerialNumber: xhci-hcd.0.auto
    [   17.802499] hub 2-0:1.0: USB hub found
    [   17.806263] hub 2-0:1.0: 1 port detected
    [   17.810592] UBI: attaching mtd2 to ubi0
    [   19.933095] UBI: scanning is finished
    [   19.944855] UBI warning: print_rsvd_warning: cannot reserve enough PEBs for bad PEB handling, reserved 40, need 76
    [   19.957109] UBI: attached mtd2 (name "ubifs", size 510 MiB) to ubi0
    [   19.963407] UBI: PEB size: 131072 bytes (128 KiB), LEB size: 126976 bytes
    [   19.970755] UBI: min./max. I/O unit sizes: 2048/2048, sub-page size 2048
    [   19.977476] UBI: VID header offset: 2048 (aligned 2048), data offset: 4096
    [   19.984382] UBI: good PEBs: 4080, bad PEBs: 4, corrupted PEBs: 0
    [   19.990412] UBI: user volume: 2, internal volumes: 1, max. volumes count: 128
    [   19.997571] UBI: max/mean erase counter: 3/1, WL threshold: 4096, image sequence number: 461878562
    [   20.006566] UBI: available PEBs: 0, total reserved PEBs: 4080, PEBs reserved for bad PEB handling: 40
    [   20.016346] UBI: background thread "ubi_bgt0d" started, PID 1311
    [   20.086259] UBIFS: mounted UBI device 0, volume 1, name "rootfs", R/O mode
    [   20.093165] UBIFS: LEB size: 126976 bytes (124 KiB), min./max. I/O unit sizes: 2048 bytes/2048 bytes
    [   20.102336] UBIFS: FS size: 258650112 bytes (246 MiB, 2037 LEBs), journal size 9023488 bytes (8 MiB, 72 LEBs)
    [   20.112290] UBIFS: reserved for root: 0 bytes (0 KiB)
    [   20.117357] UBIFS: media format: w4/r0 (latest is w4/r0), UUID 1C906995-6A08-427C-AD3E-2F47B6C0A0F5, small LPT model
    [   20.129258] VFS: Mounted root (ubifs filesystem) readonly on device 0:12.
    [   20.137663] devtmpfs: mounted
    [   20.140792] Freeing unused kernel memory: 328K (c0752000 - c07a4000)
    INIT: version 2.88 booting
    Starting udev
    [   20.650718] udevd[1341]: starting version 182
    [   21.158815] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    [   24.358755] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    [   27.558748] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    [   30.758758] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    [   33.958747] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    [   37.158754] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    [   40.358755] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    [   43.558747] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    [   46.758757] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    [   49.958747] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    udevd[1341]: worker [1347] timeout, kill it
    
    udevd[1341]: seq 1372 '/devices/soc.1/2680000.usb/2690000.dwc3/xhci-hcd.0.auto/usb1' killed
    
    Starting Bootlog daemon: bootlogd.
    [   53.158803] hub 1-0:1.0: connect-debounce failed, port 1 disabled
    rm: cannot remove '///etc/resolv.conf': Read-only file system
    ln: cannot remove '///etc/resolv.conf': Read-only file system
    Configuring network interfaces... ifup: don't seem to have all the variables for eth0/inet
    done.
    net.ipv4.conf.default.rp_filter = 1
    net.ipv4.conf.all.rp_filter = 1
    Thu Apr 24 06:53:00 UTC 2014
    INIT: Entering runlevel: 5
    Starting system message bus: dbus.
    Starting Dropbear SSH server: Will output 1024 bit rsa secret key to '/var/lib/dropbear/dropbear_rsa_host_key'
    Generating key, this may take a while...
    
    ...
    Removed due to the inclusion of customer information.
    ...
    
    dropbear.
    Starting mpmsrv daemon.
    Starting softhsm daemon.
    Starting tiipclad daemon.
    
    ...
    Removed due to the inclusion of customer information.
    ...
    
    udevd[1341]: worker [1347] terminated by signal 9 (Killed)
    

    Are the affected and unaffected boards using the same device tree?

    The same device tree is being used.

    With USB disabled, does the delay remain absent during a long-duration test?

    I have confirmed that the IPC latency—which consistently occurred when the USB controller was enabled—does not recur when the USB controller is disabled. I plan to conduct long-duration testing separately.

    Best regards,
    O.H

  • Hi O.H,

    Thank you for the answers. Since USB is not planned for use, we recommend disabling it in the device tree so that the software configuration matches the board. There are two points from your log worth highlighting:

    • The debounce failure is not a one-off event. It occurs in intervals and is still recurring when the log ends. This indicates that USB register access is happening repeatedly on the same internal bus used by the IPC registers.
    • The driver also reports "no usb2 phy configured" before the controller comes up, which is consistent with an unpowered PHY.

    There are no bus errors or timeouts reported anywhere in the log, indicating that the accesses are completing rather than failing. The result of the long-duration test with USB disabled would help confirm whether the issue is fully resolved. Could you please share the results of the long-duration test once it is completed?

    Please refer the section 2.2.8 USB Power Supplies from the Hardware Design Guide for KeyStone II Devices.

    Regards,

    Betsy Varughese

  • Hi Betsy Varughese,

    Thank you for your supports. Customer feedback below.

    We completed a 24-hour continuous test on a previously affected unit with the USB controller disabled in the Linux device tree. No IPC delay occurred during the test.

    This result strengthens the observed correlation between USB controller activity and the IPC delay. However, the delay occurs only on some SoCs, even though the affected and unaffected units use the same board revision and software configuration.

    Could TI please clarify the following?

    1. Could silicon process variation, silicon revision, or production lot differences explain why only some devices exhibit the IPC delay?

    2. Are there any known errata or internal cases involving:

      • An enabled USB controller with unpowered USB PHY supplies,
      • Repeated USB access or connect-debounce retries,
      • Internal bus contention affecting IPCGR access, or
      • A delayed IPC event or IPCGR read-back?
    3. Are there any known TI E2E cases in which USB activity or an unpowered USB PHY affected IPC or other peripherals sharing the same internal bus?

    4. For a product that does not use USB and has the USB supply pins tied to GND, can TI confirm that disabling the USB controller in the device tree is sufficient as a permanent corrective action?

    5. Is there any known aging or long-term reliability mechanism that could cause currently unaffected devices to develop this issue later?

    Regarding the board version, our understanding is that the affected units and the units operating normally are the same hardware revision.
    Unfortunately, it is currently difficult for us to determine the silicon (chip) revision of the affected devices.

    If silicon revision information is required for further investigation, we can ask the customer to identify and provide the silicon revision details.

    Best regards,
    O.H

  • Hi O.H,

    Could you please ask the customer to share the silicon revision/JTAGID details as well? It would be helpful.

    Regards,

    Betsy Varughese

  • Hi Betsy Varughese,

    We checked the silicon revision information on both an affected unit and an unaffected unit.

    Affected unit:

    • 0x01812000: 0x00090003
    • 0x02620018: 0xBB98102F

    Unaffected unit:

    • 0x01812000: 0x00090003
    • 0x02620018: 0xBB98102F

    The values are identical, so we do not see any apparent hardware difference based on these registers. In addition, we performed a 24-hour continuous operation test on a previously affected unit with the USB controller disabled in the Linux device tree. No IPC delay was observed during the 24-hour test.

    At this point, the most important question for us is whether devices that have never exhibited the IPC delay could potentially develop the same symptom in the future due to aging, long-term reliability effects, or any device-specific mechanism.

    This information will be very important for determining whether corrective action is required for products that have already been shipped to end customers. At present, our planned corrective action would be to disable the USB controller in the device tree, since USB is not used in this product.

    Based on the investigation results so far, does TI believe that:

    1. Disabling the USB controller in the device tree is a sufficient permanent corrective action?
    2. The issue is more likely related to the unsupported USB configuration rather than silicon degradation or a reliability concern?
    3. There is any risk that currently unaffected devices could begin exhibiting the IPC delay later in their lifetime?

    We would greatly appreciate TI's assessment, as this will directly affect our decision regarding fielded products.

    Best regards,
    O.H

  • Hi O.H,

    Let me discuss this internally and get back to you.

    Regards,

    Betsy Varughese

  • Hi O.H,

    Thank you for sharing the requested details.

    Both units read JTAGID = 0xBB98102F and C66x CorePac REVID = 0x00090003, which is silicon revision 3.1 on both (SPRZ402F, Table 2). The silicon revision is therefore not the reason for the difference between the units.

    Since USB is not used and the USB supply pins are tied to GND, it is recommend to keep the USB controller disabled in the device tree.

    The USB controller registers and the IPCGR registers (in the Boot Configuration module) are reached through the same internal configuration bridge (66AK2H06 data manual SPRS866, Figure 9-7).

    Normally, once USB is initialized and nothing is connected, there is no USB register activity, so the bridge is idle and IPC is not affected. On the affected unit, however, the boot log shows "connect-debounce failed, port 1 disabled" repeating every 3.2 s. This means port 1 keeps reporting a connection change although nothing is connected, so Linux keeps handling USB port events and accessing the USB registers. If one of these accesses takes longer than normal because the USB PHY is not powered, the IPCGR access waits behind it on the shared bridge, which delays the IPC interrupt. The continuous connection detection may come from the board, for example from how the unused USB signal pins are connected, or from unit-to-unit variation in how the unpowered USB PHY behaves. We cannot tell which from the current data, and the questions below will help us confirm it. In either case, keeping the USB controller disabled in the device tree removes the trigger.

    1. Do the unaffected units currently run with the USB controller enabled in the device tree, the same as the affected unit before the change?
    2. If yes, does the boot log (dmesg) of an unaffected unit also show "connect-debounce failed, port 1 disabled"? Please share the boot log of one unaffected unit with USB enabled.
    3. How are the remaining USB pins connected on your board?
    4. On the affected unit and on one unaffected unit, with the board powered on, could you measure the voltage of the USB supplies (VPH, VDDUSB, VP, VPTX and DVDD33)? Do both units read the same?
    5. Is there any known difference between the affected and unaffected units, such as a different component lot, or a different assembly batch?

    Regards,
    Betsy Varughese

  • Hi Betsy Varughese,

    Sorry for late reply. Customer feedback below. 

    Since the schematic and boot log contain customer-confidential information, I have shared them with you separately via private chat rather than posting them in this public thread.

    ====================

    1. Do the unaffected units currently run with the USB controller enabled in the device tree, the same as the affected unit before the change?

    Yes, the USB controller is also enabled on the unaffected units.

    2. If yes, does the boot log (dmesg) of an unaffected unit also show "connect-debounce failed, port 1 disabled"? Please share the boot log of one unaffected unit with USB enabled.

    We performed testing for several days, but did not observe this message in the boot log of the unaffected unit. Please refer to the attached unit_ok_log.txt.

    3. How are the remaining USB pins connected on your board?

    Please refer to the attached schematic.pdf. The components enclosed by the dashed lines in the schematic are not populated.

    4. On the affected unit and on one unaffected unit, with the board powered on, could you measure the voltage of the USB supplies (VPH, VDDUSB, VP, VPTX and DVDD33)? Do both units read the same?

    The affected unit is from the mass-production shipment, while the unaffected unit available to us is from a development lot. For VPH, VDDUSB, VP, and VPTX, we measured the voltage at the resistor locations directly connected to these pins. All of these voltages measured 0 V on both units.

    We were unable to find an accessible measurement point for DVDD33. However, since the associated power supply IC and related components are not populated, we believe DVDD33 is also at 0 V.

    The measured values are the same on both the affected and unaffected units.

    5. Is there any known difference between the affected and unaffected units, such as a different component lot, or a different assembly batch?

    The unaffected unit currently available to us is from a development lot, so its production lot is different from that of the affected unit.

    However, among the mass-produced units, there are both unaffected units and units that exhibit the issue. Although we have not checked every unit, we understand that both affected and unaffected units exist within the same production lot.

    So far, the issue has been observed on approximately three mass-produced units. Regarding the lot numbers of the SoCs purchased from distributor, two of the three affected units use SoCs from the same lot. Of these three affected units, two use SoCs purchased in 2023, while one uses an SoC purchased in 2025.

    Best regards,
    O.H

  • Hi O.H,

    Can you shared the files mentioned above?

    Regards,

    Betsy Varughese

  • Hi Betsy Varughese,

    I have sent a friend request; please accept it.

    Best regards,
    O.H

  • Hi O.H ,

    Sure.

    If USB is not being used, the current recommendation is to connect VDDUSB, VP, VPH, VPTX, and DVDD33 to ground (refer to Section 2.2.8 - USB Power Supplies - https://www.ti.com/lit/an/sprabv0/sprabv0.pdf). Since USB is not planned for use, we also recommend disabling it in the device tree to ensure the software configuration is aligned with the board design.

    You can also refer the previous E2E post similar to this:

    1.  66AK2H14: power-up sequence 

    2.  66AK2H12 hardware guide and power pins 

    Regards,

    Betsy Varughese

  • Hi O.H,

    Thank you for the schematic and the boot logs.

    The schematic follows the KeyStone II hardware design guide for an unused USB interface: the USB supply pins are tied to GND (SPRABV0, Sections 2.2.8 and 2.3.5), and the unused USB signals are left unconnected (SPRABV0, Section 6.13.7).

    1. Permanent corrective action: Yes, For a product that does not use USB and has the USB supply pins tied to GND, keeping the USB controller disabled in the device tree (usb@2680000 and usb_phy@2620738) on all units is the recommended corrective action. USB cannot operate without its supplies (SPRABV0, Section 6.13.2), and with the controller disabled, Linux does not access the USB registers.

    2. Is the cause the USB configuration rather than silicon degradation? Yes. The evidence points to the USB configuration, not to silicon degradation.

    3. Could unaffected units develop the delay later? With USB disabled in the device tree, no. Disabling USB removes the cause on every unit: Linux no longer runs the USB controller, so the behaviour of the unpowered USB PHY no longer matters, now or in the future.
    If USB were left enabled, we could not give this assurance. On every unit the USB PHY is unpowered, and whether its port stays quiet depends on an undefined signal level that is not a specified or characterized parameter. This is why we recommend disabling USB on all units, including the units that do not show the delay today.

    If USB is needed in the future, please power and connect the USB interface as described in SPRABV0 (Sections 2.2.8, 2.3.5 and 6.13) and follow the power-up sequence in SPRS866G (Table 11-2 and Figure 11-1) before enabling the USB nodes in the device tree.

    Regards,

    Betsy Varughese

  • Hi Betsy Varughese,

    Sorry for late reply, and thank you for your kind supports. We understood. If I have any further questions, I will start a new thread.

    Best reards,
    O.H