Skip to contentSkip to conversation

habanalabs: replace GFP_ATOMIC with GFP_KERNEL

bug report Dan Carpenter 2 participants 2 messages Aug 13, 2021 – Aug 14, 2021 Kernel janitors

lore
Share & tools
lore
#0 lore raw

Hello Ofir Bitton,

The patch d5eb8373b2ce: "habanalabs: replace GFP_ATOMIC with GFP_KERNEL" from Feb 14, 2021, leads to the following Smatch static checker warning:

drivers/misc/habanalabs/common/command_buffer.c:318 hl_cb_create()
warn: sleeping in atomic context

drivers/misc/habanalabs/common/command_buffer.c

    311 		if (rc)
    312 			goto release_cb;
    313 	}
    314 
    315 	spin_lock(&mgr->cb_lock);
                          ^^^^^^^^^^^^^
    316 	rc = idr_alloc(&mgr->cb_handles, cb, 1, 0, GFP_ATOMIC);
    317 	if (rc < 0)
--> 318 		rc = idr_alloc(&mgr->cb_handles, cb, 1, 0, GFP_KERNEL);
                                                                   ^^^^^^^^^^

This falls back to GFP_KERNEL if GFP_ATOMIC fails. I don't understand that. If it's at all a performance improvement to call idr_alloc() with ATOMIC first, then the correct thing to do is to fix idr_alloc() instead of working around it in the caller.

    319 	spin_unlock(&mgr->cb_lock);
    320 

The other warning is also idr_alloc() but holding a different spin lock.

drivers/misc/habanalabs/common/memory.c:126 alloc_device_memory()
warn: sleeping in atomic context

   124  
   125          spin_lock(&vm->idr_lock);
                ^^^^^^^^^^^^^^^^^^^^^^^^
   126          handle = idr_alloc(&vm->phys_pg_pack_handles, phys_pg_pack, 1, 0,
   127                                  GFP_KERNEL);
                                        ^^^^^^^^^^
   128          spin_unlock(&vm->idr_lock);
   129  
   130          if (handle < 0) {
   131                  dev_err(hdev->dev, "Failed to get handle for page\n");
   132                  rc = -EFAULT;
   133                  goto idr_err;
   134          }

regards,
dan carpenter

#1 lore raw
Quoted 44 lines · Dan Carpenter

[???? ???? ???? ???????? ????? ?????? ?- [email protected]. ??? ???? ???? ???? ?- http://aka.ms/LearnAboutSenderIdentification.]

Hello Ofir Bitton,

The patch d5eb8373b2ce: "habanalabs: replace GFP_ATOMIC with GFP_KERNEL" from Feb 14, 2021, leads to the following Smatch static checker warning:

drivers/misc/habanalabs/common/command_buffer.c:318 hl_cb_create()
warn: sleeping in atomic context

drivers/misc/habanalabs/common/command_buffer.c

    311                 if (rc)
    312                         goto release_cb;
    313         }
    314
    315         spin_lock(&mgr->cb_lock);
                          ^^^^^^^^^^^^^
    316         rc = idr_alloc(&mgr->cb_handles, cb, 1, 0, GFP_ATOMIC);
    317         if (rc < 0)
--> 318                 rc = idr_alloc(&mgr->cb_handles, cb, 1, 0, GFP_KERNEL);
                                                                   ^^^^^^^^^^ This falls back to GFP_KERNEL if GFP_ATOMIC fails.  I don't understand that.  If it's at all a performance improvement to call idr_alloc() with ATOMIC first, then the correct thing to do is to fix idr_alloc() instead of working around it in the caller.

    319         spin_unlock(&mgr->cb_lock);
    320

The other warning is also idr_alloc() but holding a different spin lock.

drivers/misc/habanalabs/common/memory.c:126 alloc_device_memory()
warn: sleeping in atomic context

   124
   125          spin_lock(&vm->idr_lock);
                ^^^^^^^^^^^^^^^^^^^^^^^^
   126          handle = idr_alloc(&vm->phys_pg_pack_handles, phys_pg_pack, 1, 0,
   127                                  GFP_KERNEL);
                                        ^^^^^^^^^^
   128          spin_unlock(&vm->idr_lock);
   129
   130          if (handle < 0) {
   131                  dev_err(hdev->dev, "Failed to get handle for page\n");
   132                  rc = -EFAULT;
   133                  goto idr_err;
   134          }
Quoted 2 lines

regards,
dan carpenter

Thanks for noticing I will handle it.
Ofir