Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S967514Ab3DRJgj (ORCPT ); Thu, 18 Apr 2013 05:36:39 -0400 Received: from fgwmail5.fujitsu.co.jp ([192.51.44.35]:36933 "EHLO fgwmail5.fujitsu.co.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S967359Ab3DRJfH (ORCPT ); Thu, 18 Apr 2013 05:35:07 -0400 From: HATAYAMA Daisuke Subject: [PATCH v4 5/8] vmcore: copy ELF note segments in the 2nd kernel per page vmcore objects To: vgoyal@redhat.com, ebiederm@xmission.com, akpm@linux-foundation.org Cc: cpw@sgi.com, kumagai-atsushi@mxc.nes.nec.co.jp, lisa.mitchell@hp.com, kexec@lists.infradead.org, linux-kernel@vger.kernel.org, zhangyanfei@cn.fujitsu.com, jingbai.ma@hp.com Date: Sat, 13 Apr 2013 09:21:33 +0900 Message-ID: <20130413002133.18245.91528.stgit@localhost6.localdomain6> In-Reply-To: <20130413002000.18245.21513.stgit@localhost6.localdomain6> References: <20130413002000.18245.21513.stgit@localhost6.localdomain6> User-Agent: StGIT/0.14.3 MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4542 Lines: 145 Copy ELF note segmnets in buffer on the 2nd kernel and the buffer is allocated on page boundary for the purpose of mmap. The reason why we don't allocate ELF note segments in the 1st kernel (old memory) on page boundary is to keep backward compatibility for old kernels, and one more reason is that if doing so, we waste not a little memory due to round-up operation to fit the memory to page boundary since most of the buffers are in per-cpu area. ELF notes are per-cpu, so total size of ELF note segments increases according to the number of CPUs. The current maximum number of CPUs on x86_64 is 5192, and there's already system with 4192 CPUs in SGI, where total size amounts to 1MB. This can be larger in the neare futrue or possibly even now on another architecture. Thus, to avoid the case where memory allocation for large block fails, we allocate vmcore objects per pages. Note that we cannot use remap_vmalloc_range here since the vma passed to it as the first argument needs to cover a full range of memory. We need to be able to specify where to map in pages, but we cannot do it by remap_vmalloc_range. Signed-off-by: HATAYAMA Daisuke --- fs/proc/vmcore.c | 71 ++++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 files changed, 69 insertions(+), 2 deletions(-) diff --git a/fs/proc/vmcore.c b/fs/proc/vmcore.c index 131d8fa..e27da40 100644 --- a/fs/proc/vmcore.c +++ b/fs/proc/vmcore.c @@ -200,7 +200,62 @@ static int __init vmcore_add(struct list_head *vc_list, u64 paddr, u64 size) return 0; } +struct vmcore_per_unit_state { + struct vmcore *last; + u64 pos; + u64 unit; + struct list_head *vc_list; +}; + +/* + * Assumptions: + * - given physical addresses are all exclusive; don't check exclusiveness. + */ +static int __init vmcore_add_per_unit(struct vmcore_per_unit_state *state, + u64 paddr, u64 size) +{ + u64 offset = paddr, remaining_bytes = size; + int rc; + + while (remaining_bytes > 0) { + u64 read_bytes; + + BUG_ON(state->pos > state->unit); + + if (!state->last || state->pos == state->unit) { + struct vmcore *new; + + new = get_new_element(); + if (!new) + return -ENOMEM; + new->flag = MEM_TYPE_CURRENT_KERNEL; + new->size = PAGE_SIZE; + new->buf = (char *)get_zeroed_page(GFP_KERNEL); + if (!new->buf) { + kfree(new); + return -ENOMEM; + } + list_add_tail(&new->list, state->vc_list); + state->last = new; + state->pos = 0; + } + + read_bytes = min(remaining_bytes, state->unit - state->pos); + + rc = read_from_oldmem(state->last->buf + state->pos, read_bytes, + &offset, 0); + if (rc < 0) + return rc; + + state->pos += read_bytes; + remaining_bytes -= read_bytes; + } + + return 0; +} + static u64 __init get_vmcore_size_elf64(char *elfptr, size_t elfsz) + { int i; u64 size; @@ -244,6 +299,12 @@ static int __init merge_note_headers_elf64(char *elfptr, size_t *elfsz, Elf64_Phdr phdr, *phdr_ptr; Elf64_Nhdr *nhdr_ptr; u64 phdr_sz = 0, note_off; + struct vmcore_per_unit_state state = { + .last = NULL, + .pos = 0, + .unit = PAGE_SIZE, + .vc_list = vc_list, + }; ehdr_ptr = (Elf64_Ehdr *)elfptr; phdr_ptr = (Elf64_Phdr*)(elfptr + sizeof(Elf64_Ehdr)); @@ -276,7 +337,7 @@ static int __init merge_note_headers_elf64(char *elfptr, size_t *elfsz, } /* Add this contiguous chunk of notes section to vmcore list.*/ - if (vmcore_add(vc_list, phdr_ptr->p_offset, real_sz)) { + if (vmcore_add_per_unit(&state, phdr_ptr->p_offset, real_sz)) { kfree(notes_section); return -ENOMEM; } @@ -323,6 +384,12 @@ static int __init merge_note_headers_elf32(char *elfptr, size_t *elfsz, Elf32_Phdr phdr, *phdr_ptr; Elf32_Nhdr *nhdr_ptr; u64 phdr_sz = 0, note_off; + struct vmcore_per_unit_state state = { + .last = NULL, + .pos = 0, + .unit = PAGE_SIZE, + .vc_list = vc_list, + }; ehdr_ptr = (Elf32_Ehdr *)elfptr; phdr_ptr = (Elf32_Phdr*)(elfptr + sizeof(Elf32_Ehdr)); @@ -355,7 +422,7 @@ static int __init merge_note_headers_elf32(char *elfptr, size_t *elfsz, } /* Add this contiguous chunk of notes section to vmcore list.*/ - if (vmcore_add(vc_list, phdr_ptr->p_offset, real_sz)) { + if (vmcore_add_per_unit(&state, phdr_ptr->p_offset, real_sz)) { kfree(notes_section); return -ENOMEM; } -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/