On 04/02/25 10:53AM, Jonathan Cameron wrote:
On Mon,  3 Feb 2025 11:29:49 +0530
Arpit Kumar <arpit1.ku...@samsung.com> wrote:

Add some description of what is being added here.

Key issue in here is that clearing the CEL doesn't make
sense. It is a description of what the device supports, there
is no state to clear in it.  To add this command you need
to pick a different log.

Jonathan

Thanks for the review Jonathan, will modify the code accordingly in V2 patch.


Signed-off-by: Arpit Kumar <arpit1.ku...@samsung.com>
Reviewed-by: Alok Rathore <alok.rath...@samsung.com>
Reviewed-by: Krishna Kanth Reddy <krish.re...@samsung.com>
---
 hw/cxl/cxl-mailbox-utils.c | 36 ++++++++++++++++++++++++++++++++++++
 1 file changed, 36 insertions(+)

diff --git a/hw/cxl/cxl-mailbox-utils.c b/hw/cxl/cxl-mailbox-utils.c
index 3d66a425a9..5fd7f850c4 100644
--- a/hw/cxl/cxl-mailbox-utils.c
+++ b/hw/cxl/cxl-mailbox-utils.c
@@ -77,6 +77,7 @@ enum {
         #define GET_SUPPORTED 0x0
         #define GET_LOG       0x1
         #define GET_LOG_CAPABILITIES   0x2
+        #define CLEAR_LOG     0x3
     FEATURES    = 0x05,
         #define GET_SUPPORTED 0x0
         #define GET_FEATURE   0x1
@@ -1115,6 +1116,39 @@ static CXLRetCode cmd_logs_get_log_capabilities(const 
struct cxl_cmd *cmd,
     return CXL_MBOX_SUCCESS;
 }

+/* CXL r3.1 Section 8.2.9.5.4: Clear Log (Opcode 0403h) */
+static CXLRetCode cmd_logs_clear_log(const struct cxl_cmd *cmd,
+                                     uint8_t *payload_in,
+                                     size_t len_in,
+                                     uint8_t *payload_out,
+                                     size_t *len_out,
+                                     CXLCCI *cci)
+{
+    int32_t cap_id;
+    struct {
+        QemuUUID uuid;
+    } QEMU_PACKED QEMU_ALIGNED(8) * clear_log = (void *)payload_in;
+
+    cap_id = valid_log_check(&clear_log->uuid, cci);
+    if (cap_id == -1) {
+        return CXL_MBOX_INVALID_LOG;
+    }

Follow on from previous patch, if this returns the cap pointer,
the following code wont have to index the array and should end up simpler.

okay

+
+    if (cci->supported_log_cap[cap_id].param_flags.clear_log_supported) {
I would flip this.
   if (!(cap->param_flags & PARAM_FLAG_CLEAR_LOG_SUPPORTED)) {
       return CXL_MBOX_UNSUPPORTED;
   }


+        switch (cap_id) {
+        case CEL:

So if we return the cap as suggested, it will have to reference what it is
or provide a callback (which might be cleaner as this grows).

However, what does clearly the command effects log mean?
This makes no sense.  So if you want to implement clear_log you
need to implement a different log to clear.

okay

+            memset(cci->cel_log, 0, (1 << 16) * sizeof(struct cel_log));
+            cci->cel_size = 0;
+            break;
+        default:
+            return CXL_MBOX_UNSUPPORTED;
+        }
+    } else {
+        return CXL_MBOX_UNSUPPORTED;
+    }
+    return CXL_MBOX_SUCCESS;
+}
+
 /* CXL r3.1 section 8.2.9.6: Features */
 /*
  * Get Supported Features output payload
@@ -2882,6 +2916,8 @@ static const struct cxl_cmd cxl_cmd_set[256][256] = {
     [LOGS][GET_LOG] = { "LOGS_GET_LOG", cmd_logs_get_log, 0x18, 0 },
     [LOGS][GET_LOG_CAPABILITIES] = { "LOGS_GET_LOG_CAPABILITIES",
                                      cmd_logs_get_log_capabilities, 0x10, 0 },
+    [LOGS][CLEAR_LOG] = { "LOGS_CLEAR_LOG", cmd_logs_clear_log, 0x10,
+                          CXL_MBOX_IMMEDIATE_LOG_CHANGE},
     [FEATURES][GET_SUPPORTED] = { "FEATURES_GET_SUPPORTED",
                                   cmd_features_get_supported, 0x8, 0 },
     [FEATURES][GET_FEATURE] = { "FEATURES_GET_FEATURE",



Reply via email to