Recovering from uncorrected memory errors

1. A method of recovering uncorrected memory errors, the method comprising:

receiving, at an operating system, a corrected error (CE) associated with a first memory page;

marking the corrected error in a page table entry describing the first memory page;

migrating the first memory page, by the operating system, to a second memory page, based on the received corrected error, wherein the second memory page is error-free, wherein the first memory page is a data page, wherein the data page includes program variables that are mapped into memory from a data segment in an executable program file;

taking no action wherein the first memory page is a text page; and

flushing the first memory page to a file on non-transient storage, wherein the first memory page is a file page whose contents are mapped into memory from the file that resides on non-transient storage.

2. The method of claim 1 , wherein flushing the first memory page further comprises:

storing a plurality of first access permissions associated with the first memory page in the page table entry describing the first memory page wherein the first memory page is the file page;

marking the first memory page as having a read-only access permission; and

writing a contents of the first memory page to the file on non-transient storage, based on a status indicating that the contents of the first memory page are altered.

3. The method of claim 1 , wherein migrating the first memory page further comprises:

copying the contents of the first memory page to the second memory page, wherein the first memory page is the data page and the second memory page is error-free;

removing an address reference to the first memory page from a plurality of tables describing an address space;

adding an address reference to the second memory page to the plurality of tables describing the address space; and

adding the first memory page to an operating system queue of free memory pages having one or more corrected errors.

4. The method of claim 1 , further comprising:

discarding the contents of the first memory page;

re-fetching, by the operating system, the contents of the first memory page from non-transient storage based on receiving an uncorrected error (UE) associated with the first memory page and based on the first memory page receiving the CE prior to receiving the UE; and

terminating abnormally, based on the first memory page receiving the UE without receiving the prior CE.

5. The method of claim 1 , further comprising:

allocating a free memory page from the operating system queue of free memory pages having one or more corrected errors, when an operating system queue of error-free pages is empty and a request for allocating the free memory page is for a read-only page wherein the read-only page includes a text page and a file page.