Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 1 | TODO |
| 2 | ==== |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 3 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 4 | API |
| 5 | === |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 6 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 7 | DMA Resource Allocation incomplete |
| 8 | ---------------------------------- |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 9 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 10 | The current DMA resource Allocation provides no means of selecting the |
| 11 | suitability of a DMA controller based on it's supported modes of operation, as |
| 12 | opposed to the resource allocation mechanisms for master and slave windows: |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 13 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 14 | struct vme_resource *vme_request_dma(struct device *dev); |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 15 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 16 | As opposed to: |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 17 | |
| 18 | struct vme_resource * vme_master_request(struct device *dev, |
| 19 | vme_address_t aspace, vme_cycle_t cycle, vme_width_t width); |
| 20 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 21 | The TSI148 can perform, VME-to-PCI, PCI-to-VME, PATTERN-to-VME, PATTERN-to-PCI, |
| 22 | VME-to-VME and PCI-to-PCI transfers. The CA91C142 can only provide VME-to-PCI |
| 23 | and PCI-to-VME. |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 24 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 25 | Add a mechanism to select a VME controller based on source/target type, |
| 26 | required aspace, cycle and width requirements. |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 27 | |
| 28 | |
| 29 | Master window broadcast select mask |
| 30 | ----------------------------------- |
| 31 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 32 | API currently provides no method to set or get Broadcast Select mask. Suggest |
| 33 | somthing like: |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 34 | |
| 35 | int vme_master_bmsk_set (struct vme_resource *res, int mask); |
| 36 | int vme_master_bmsk_get (struct vme_resource *res, int *mask); |
| 37 | |
| 38 | |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 39 | Interrupt Generation |
| 40 | -------------------- |
| 41 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 42 | Add optional timeout when waiting for an IACK. |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 43 | |
| 44 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 45 | CR/CSR Buffer |
| 46 | ------------- |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 47 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 48 | The VME API provides no functions to access the buffer mapped into the CR/CSR |
| 49 | space. |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 50 | |
| 51 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 52 | Mailboxes |
| 53 | --------- |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 54 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 55 | Whilst not part of the VME specification, they are provided by a number of |
| 56 | chips. They are currently not supported at all by the API. |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 57 | |
| 58 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 59 | Core |
| 60 | ==== |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 61 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 62 | - Rename vme_master_resource's "pci_resource" to be bus agnostic. |
| 63 | - Improve generic sanity checks (Such as does an offset and size fit within a |
| 64 | window and parameter checking). |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 65 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 66 | Bridge Support |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 67 | ============== |
| 68 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 69 | Tempe (tsi148) |
| 70 | -------------- |
Martyn Welch | c7ef5da | 2009-07-31 14:07:53 +0100 | [diff] [blame] | 71 | |
Martyn Welch | bf39f9a | 2009-08-28 11:28:56 +0100 | [diff] [blame] | 72 | - Driver can currently only support a single bridge. |
| 73 | - 2eSST Broadcast mode. |
| 74 | - Mailboxes unsupported. |
| 75 | - Improve error detection. |
| 76 | - Control of prefetch size, threshold. |
| 77 | - Arbiter control |
| 78 | - Requestor control |
| 79 | |
| 80 | Universe II (ca91c142) |
| 81 | ---------------------- |
| 82 | |
| 83 | - Driver can currently only support a single bridge. |
| 84 | - DMA unsupported. |
| 85 | - RMW transactions unsupported. |
| 86 | - Location Monitors unsupported. |
| 87 | - Mailboxes unsupported. |
| 88 | - Error Detection. |
| 89 | - Control of prefetch size, threshold. |
| 90 | - Arbiter control |
| 91 | - Requestor control |
| 92 | - Slot detection |
| 93 | |
| 94 | Universe I (ca91x042) |
| 95 | --------------------- |
| 96 | |
| 97 | Currently completely unsupported. |
| 98 | |