Re: ath11k: question about memory-region
Vasanthakumar Thiagarajan <[email protected]>
| Newsgroups | org.infradead.lists.ath11k,org.kernel.vger.linux-wireless |
|---|---|
| Message-ID | <[email protected]> |
On 9/3/2025 2:01 PM, Alexander Wilhelm wrote: > Hello devs, > > I have a question regarding the use of memory-region on a platform without CMA > support. Since CMA is not available, the driver allocates many smaller memory > chunks instead. > > To evaluate an alternative, I defined a node within reserved-memory in the > device tree. However, this led to a firmware crash: > > ath11k_pci 0001:01:00.0: BAR 0: assigned [mem 0xc00000000-0xc001fffff 64bit] > ath11k_pci 0001:01:00.0: boot pci_mem 0x10b272d0 > ath11k_pci 0001:01:00.0: boot pci probe 17cb:1104 17cb:1104 > ath11k_pci 0001:01:00.0: pci request one msi vector > ath11k_pci 0001:01:00.0: MSI vectors: 1 > ath11k_pci 0001:01:00.0: pci msi base data is 0 > ath11k_pci 0001:01:00.0: qcn9074 hw1.0 > ath11k_pci 0001:01:00.0: FW memory mode: 0 > ath11k_pci 0001:01:00.0: boot failed to load firmware-2.bin: -2 > ath11k_pci 0001:01:00.0: boot using fw api 1 > ath11k_pci 0001:01:00.0: pci msi assignment MHI num_vectors 3 user_base_data 0 base_vector 0 > ath11k_pci 0001:01:00.0: pci num_vectors 3 base_vector 0 > ath11k_pci 0001:01:00.0: pci msi assignment CE num_vectors 1 user_base_data 0 base_vector 0 > ath11k_pci 0001:01:00.0: pci msi assignment DP num_vectors 1 user_base_data 0 base_vector 0 > ath11k_pci 0001:01:00.0: pci irq 18 group 0 > ath11k_pci 0001:01:00.0: pci irq 18 group 1 > ath11k_pci 0001:01:00.0: pci irq 18 group 2 > ath11k_pci 0001:01:00.0: pci irq 18 group 3 > ath11k_pci 0001:01:00.0: pci irq 18 group 4 > ath11k_pci 0001:01:00.0: pci irq 18 group 5 > ath11k_pci 0001:01:00.0: pci irq 18 group 6 > ath11k_pci 0001:01:00.0: pci irq 18 group 7 > ath11k_pci 0001:01:00.0: pci after request_irq msi_ep_base_data 0 > ath11k_pci 0001:01:00.0: pci ltssm 0x111 > ath11k_pci 0001:01:00.0: pci pcie_hot_rst 0x11 > ath11k_pci 0001:01:00.0: pci pcie_q6_cookie_addr 0x0 > ath11k_pci 0001:01:00.0: pci wlaon_warm_sw_entry 0x0 > ath11k_pci 0001:01:00.0: pci wlaon_warm_sw_entry 0x0 > ath11k_pci 0001:01:00.0: pci soc reset cause 0 > ath11k_pci 0001:01:00.0: pci mhistatus 0xff04 > ath11k_pci 0001:01:00.0: pci link_ctl 0x0000 L0s 0 L1 0 > ath11k_pci 0001:01:00.0: pci pci reg 0x3164 instance_id 0x11 read val 0x11 > ath11k_pci 0001:01:00.0: boot notify status reason UNKNOWN > ath11k_pci 0001:01:00.0: boot notify status reason MHI_CB_EE_MISSION_MODE > ath11k_pci 0001:01:00.0: qmi wifi fw qmi service connected > ath11k_pci 0001:01:00.0: qmi indication register request > ath11k_pci 0001:01:00.0: qmi host cap request > ath11k_pci 0001:01:00.0: qmi firmware request memory request > ath11k_pci 0001:01:00.0: qmi mem seg type 1 size 36700160 > ath11k_pci 0001:01:00.0: qmi mem seg type 4 size 3784704 > ath11k_pci 0001:01:00.0: qmi mem seg type 3 size 1048576 > ath11k_pci 0001:01:00.0: qmi mem seg type 9 size 8388608 > ath11k_pci 0001:01:00.0: qmi mem seg type 10 size 8192 > ath11k_pci 0001:01:00.0: qmi ignore invalid mem req type 3 > ath11k_pci 0001:01:00.0: qmi ignore invalid mem req type 9 > ath11k_pci 0001:01:00.0: qmi ignore invalid mem req type 10 > ath11k_pci 0001:01:00.0: qmi req mem_seg[0] 0x0000000040000000 36700160 1 > ath11k_pci 0001:01:00.0: qmi req mem_seg[1] 0x0000000000000000 3784704 4 > ath11k_pci 0001:01:00.0: qmi respond memory request delayed 0 > ath11k_pci 0001:01:00.0: boot notify status reason MHI_CB_EE_RDDM > ath11k_pci 0001:01:00.0: firmware crashed: MHI_CB_EE_RDDM > ath11k_pci 0001:01:00.0: ignore reset dev flags 0x2000 > ath11k_pci 0001:01:00.0: failed to wait qmi memory request: -110 > ath11k_pci 0001:01:00.0: qmi failed to respond fw mem req: -110 > > Why is `CALDB_MEM_REGION_TYPE` always set to physical address 0? Coldboot cal is disabled for qcn9074. > Could the crash be caused by platform-specific constraints and not the firmware/driver? Hard to say. At least from WLAN side there is no known issues when using reserved-memory. Can you share your reserved memory node? > Does it make sense at all to use reserved-memory instead of small memory chunks, for > example in terms of performance? There will not be any visible performance differences between reserved-memory and small memory chunks. Vasanth