Bug 204286
| Summary: | 4/4GB split issue in is_prefetch() | ||||||
|---|---|---|---|---|---|---|---|
| Product: | Red Hat Enterprise Linux 4 | Reporter: | Vasily Averin <vvs> | ||||
| Component: | kernel | Assignee: | Peter Zijlstra <pzijlstr> | ||||
| Status: | CLOSED WONTFIX | QA Contact: | Brian Brock <bbrock> | ||||
| Severity: | medium | Docs Contact: | |||||
| Priority: | medium | ||||||
| Version: | 4.4 | CC: | dev, khorenko, lwang, mingo, vgoyal | ||||
| Target Milestone: | --- | ||||||
| Target Release: | --- | ||||||
| Hardware: | i686 | ||||||
| OS: | Linux | ||||||
| Whiteboard: | |||||||
| Fixed In Version: | Doc Type: | Bug Fix | |||||
| Doc Text: | Story Points: | --- | |||||
| Clone Of: | Environment: | ||||||
| Last Closed: | 2012-06-20 16:01:41 UTC | Type: | --- | ||||
| Regression: | --- | Mount Type: | --- | ||||
| Documentation: | --- | CRM: | |||||
| Verified Versions: | Category: | --- | |||||
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |||||
| Cloudforms Team: | --- | Target Upstream Version: | |||||
| Embargoed: | |||||||
| Attachments: |
|
||||||
Created attachment 135031 [details]
this patch fixes 4/4GB split issue in is_prefetch()
this patch fixes 4/4GB split issue in is_prefetch()
Signed-off-by: Kirill Korotaev <dev>
Signed-off-by: Vasily Averin <vvs>
SWsoft Virtuozzo/OpenVZ Linux kernel team
Ingo, We have found 4/4GB split related issue in is_prefetch(), could you please comment it? Thank you for submitting this issue for consideration in Red Hat Enterprise Linux. The release for which you requested us to review is now End of Life. Please See https://access.redhat.com/support/policy/updates/errata/ If you would like Red Hat to re-consider your feature request for an active release, please re-open the request via appropriate support channels and provide additional supporting details about the importance of this issue. |
SWsoft Virtuozzo/OpenVZ Linux kernel team has found 4/4GB split related issue: Recently we have investigated the cause of the node crashes on i686 kernel with 4/4Gb split patch running on AMD processor and found that it was page fault inside tcp_v4_rcv() function on prefetchnta instruction. According to Revision Guide for AMD Athlon 64 and AMD Opteron Processors, there is AMD CPU errata #91 (Software Prefetches May Report A Page Fault). Linux kernel have a workaround for this issue, do_page_fault() returns if EIP points to prefetch instruction: if (is_prefetch(regs, address, error_code)) return; To read current opcode __is_prefetch() uses __get_user() function. It works well on usual kernels however the problem is that 4/4G split patch requires to set_fs(KERNEL_DS) in this case. Otherwise kernel reads this address from userspace and does not found an according opcode, is_prefetch returns 0 and as result it leads to a kernel panic. Originally this issue was found on Virtuozzo/OpenVZ i686 enetrprise kernels, however we believe it present in Red Hat hugemem kernels too.