Re: [Qemu-devel] [PATCH 00/10] linux-user: Speed up guest space initialization on 32-bit ARM target
Ping. On Thu, 28 Dec 2017 13:08:03 -0500, Luke Shumaker wrote: > > From: Luke Shumaker > > The goal of this patchset is to fix > https://bugs.launchpad.net/qemu/+bug/1740219 > > The gist is that the current linear search for an acceptable address > range is a bad strategy when the reason we didn't get a good address > on the first try is that we've having problems mapping the commpage > for the 32-bit ARM target; especially if ASLR is disabled. > > I think that only the final patch in this patchset is actually > necessary to fix the issue; but I didn't feel comfortable writing it > without also makeing the preceding (small) changes. > > Luke Shumaker (10): > linux-user: Use #if to only call validate_guest_space for 32-bit ARM > target > linux-user: Rename validate_guest_space => init_guest_commpage > linux-user: init_guest_space: Clean up if we can't initialize the > commpage > linux-user: init_guest_space: Correctly handle guest_start in commpage > initialization > linux-user: init_guest_space: Clarify page alignment logic > linux-user: init_guest_commpage: Add a comment about size check > linux-user: init_guest_space: Clean up control flow a bit > linux-user: init_guest_space: Don't try to align if we'll reject it > linux-user: init_guest_space: Add a comment about search strategy > linux-user: init_guest_space: Try to make ARM space+commpage > continuous > > linux-user/elfload.c | 145 > --- > 1 file changed, 114 insertions(+), 31 deletions(-) > > -- > 2.15.1 > > Happy hacking, > ~ Luke Shumaker >
Re: [Qemu-devel] [PATCH 00/10] linux-user: Speed up guest space initialization on 32-bit ARM target
Ping for code review? -- Happy hacking, ~ Luke Shumaker
[Qemu-devel] [PATCH 00/10] linux-user: Speed up guest space initialization on 32-bit ARM target
From: Luke Shumaker The goal of this patchset is to fix https://bugs.launchpad.net/qemu/+bug/1740219 The gist is that the current linear search for an acceptable address range is a bad strategy when the reason we didn't get a good address on the first try is that we've having problems mapping the commpage for the 32-bit ARM target; especially if ASLR is disabled. I think that only the final patch in this patchset is actually necessary to fix the issue; but I didn't feel comfortable writing it without also makeing the preceding (small) changes. Luke Shumaker (10): linux-user: Use #if to only call validate_guest_space for 32-bit ARM target linux-user: Rename validate_guest_space => init_guest_commpage linux-user: init_guest_space: Clean up if we can't initialize the commpage linux-user: init_guest_space: Correctly handle guest_start in commpage initialization linux-user: init_guest_space: Clarify page alignment logic linux-user: init_guest_commpage: Add a comment about size check linux-user: init_guest_space: Clean up control flow a bit linux-user: init_guest_space: Don't try to align if we'll reject it linux-user: init_guest_space: Add a comment about search strategy linux-user: init_guest_space: Try to make ARM space+commpage continuous linux-user/elfload.c | 145 --- 1 file changed, 114 insertions(+), 31 deletions(-) -- 2.15.1 Happy hacking, ~ Luke Shumaker