Hi, On 3/25/07, Robert P. J. Day <rpjday@xxxxxxxxxxxxxx> wrote:
from everything i'd ever read, i was under the impression that a memory allocation might be GFP_KERNEL or GFP_ATOMIC, but not both. however, there are at least a couple calls in the kernel where both of those flags are used simultaneously. does that make sense?
If it's in the kernel, then it makes sense. The right question is, what sense it makes? http://lxr.linux.no/source/include/linux/gfp.h#L61 "/* GFP_ATOMIC means both !wait (__GFP_WAIT not set) and use emergency pool */" The GFP_ATOMIC | GFP_KERNEL results in __GFP_HIGH | __GFP_WAIT | __GFP_IO | __GFP_FS. But I understand the above comment that the use of __GFP_HIGH implies to the memory allocation code that it must ignore the __GFP_WAIT flag. Therefore I think the GFP_ATOMIC | GFP_KERNEL is equal to __GFP_HIGH | __GFP_IO | __GFP_FS. That reads: the memory allocation code is not allowed to sleep (__GFP_HIGH) (probably use emergency memory pools), is allowed to perform file system operation (__GFP_FS)[*], or is allowed to perform some input/output operation (__GFP_IO)[**]. [*] One example I can imagine is that when the system is low on memory, then the memory subsystem can try to free some dirty pages. In this case it would most probably call some file system specific function, that will be responsible for writing the page contents to the disk (or otherwise synchronize the page contents with the storage area [***]). [**] I don't have any example for this one. [***] Network file system (like NFS) would probably have to send the page through the network to the server. BlackHole -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
