When a DMA coherent pool is depleted, allocation failures may or may not get reported in the kernel log depending on the allocator. The admin does have a workaround, however, by using coherent_pool= on the kernel command line. Provide some guidance on the failure and a recommended minimum size for the pools (double the size).
Showing the changes to kernel (1 of 1 file). Show all
Unified · Split
| @@ -239,12 +239,16 @@ void *dma_alloc_from_pool(struct device *dev, size_t size, · 238 unmodified lines | ||
| 239 | 239 | } |
| 240 | 240 | |
| 241 | 241 | val = gen_pool_alloc(pool, size); |
| 242 | − if (val) { | |
| 242 | + if (likely(val)) { | |
| 243 | 243 | phys_addr_t phys = gen_pool_virt_to_phys(pool, val); |
| 244 | 244 | |
| 245 | 245 | *ret_page = pfn_to_page(__phys_to_pfn(phys)); |
| 246 | 246 | ptr = (void *)val; |
| 247 | 247 | memset(ptr, 0, size); |
| 248 | + } else { | |
| 249 | + WARN_ONCE(1, "DMA coherent pool depleted, increase size " | |
| 250 | + "(recommended min coherent_pool=%zuK)\n", | |
| 251 | + gen_pool_size(pool) >> 9); | |
| 248 | 252 | } |
| 249 | 253 | if (gen_pool_avail(pool) < atomic_pool_size) |
| 250 | 254 | schedule_work(&atomic_pool_work); |