spsc_pbuf: second large write fails with -ENOMEM after buffer is emptied due to non-normalized internal indices

Question

After a full consume of an spsc_pbuf , the buffer can become logically empty while rd_idx and wr_idx both remain at a non-zero offset. This breaks the next large allocation and causes spsc_pbuf_write() to return -ENOMEM .

Environment:

nRF Connect SDK v3.2.4

Zephyr OS v4.2.99-9673eec75908

Hardware: nRF9160

Problem: 
 After a successful spsc_pbuf_write() and spsc_pbuf_read() that empties the buffer, a second write of the same size fails with -ENOMEM even though the buffer is logically empty and has enough total capacity. 
 Steps to reproduce:

Initialize an spsc_pbuf with 4096 bytes.

spsc_pbuf_write() a 2511-byte packet — succeeds.

spsc_pbuf_read() the packet — succeeds, buffer is empty.

spsc_pbuf_write() another 2511-byte packet — fails with -ENOMEM .

Root cause analysis: 
 A 2511-byte packet uses:

2511 bytes payload

2-byte length header

padding to 4-byte alignment

total internal advance = 2516 bytes

After the read, rd_idx == wr_idx == 2516 . The buffer is empty, but the indices are not reset to 0, so the next allocation sees a non-canonical empty buffer state and fails to allocate the new packet. 
 Proposed fix: 
 In spsc_pbuf_free() , after advancing rd_idx , if the buffer is empty ( rd_idx == wr_idx ), normalize the empty state by resetting the index(es) to 0. 
 Workaround: 
 Reinitialize the buffer after each successful read: 
 int ret = spsc_pbuf_read(json_cmd_pbuf, dst, dst_len);

if (ret > 0) { 
 json_cmd_pbuf = spsc_pbuf_init(json_cmd_pbuf_mem, sizeof(json_cmd_pbuf_mem), 0); 
}