Claude review: dma-buf: heaps: allow heap to specify valid heap flags

[Claude Code Review Bot <claude-review@example.com>]

Patch Review

**Subject**: `dma-buf: heaps: allow heap to specify valid heap flags`  
**Files Modified**: `drivers/dma-buf/dma-heap.c`, `include/linux/dma-heap.h`  
**Verdict**: ❌ **REJECT** (violates design principles)

#### Technical Review

Adds per-heap flag validation to allow different heaps to accept different flags.

**API Change** (include/linux/dma-heap.h:47-48):
```c
struct dma_heap_export_info {
    const char *name;
    const struct dma_heap_ops *ops;
    void *priv;
+   unsigned long valid_heap_flags;
};
```

**Implementation** (drivers/dma-buf/dma-heap.c:127-129):
```c
-   if (heap_allocation->heap_flags & ~DMA_HEAP_VALID_HEAP_FLAGS)
+   if (heap_allocation->heap_flags & ~heap->valid_heap_flags)
        return -EINVAL;
```

**Technical Correctness**: The implementation is correct and functional.

**Design Issues** (CRITICAL):

❌ **Violates original dma-buf heaps design philosophy**
- Per maintainer John Stultz: "heap flags are supposed to be more generic allocation flags that could potentially apply to all/most of the heaps"
- This was an explicit lesson learned from ION driver, which had heap-specific flags causing API chaos
- Quote from review: "heap specific flags is something I intentionally avoided"

❌ **Slippery slope**
- Once you allow per-heap flags, every heap will add its own flags
- Leads to non-portable userspace code that must know which flags work with which heaps
- Defeats the purpose of a generic dma-buf heaps API

❌ **Alternative exists**
- Per maintainer recommendation: create separate heap with different name (e.g., "system_cc_decrypted")
- This is the idiomatic way to expose different allocation behaviors
- Keeps the flag namespace clean and generic

**Maintainer Feedback**:
```
From: John Stultz <jstultz@google.com>
"heap specific flags is something I intentionally avoided. That 
can be done if you just create a heap w/ a different name."
```

**Author Response**:
```
From: Jiri Pirko <jiri@nvidia.com>
"Okay, will send v2 with the separate heap name instead."
```

**Recommendation**: **DROP THIS PATCH** and use separate heap name approach in v2.

---

---
Generated by Claude Code Patch Reviewer