Prev: [PATCH 1/2] blackfin: ptrace: fix the unsafe usage of mm/find_vma in is_user_addr_valid()
Next: sys_personality: make sure (int)personality >= 0
From: Oleg Nesterov on 27 May 2010 16:00 Any usage of mm->context.sram_list is not safe, is_user_addr_valid() and sram_free_with_lsl/sram_free_with_lsl can race with each other. Change sram_free_with_lsl/sram_free_with_lsl to take mm->mmap_sem for writing, is_user_addr_valid() was already modified to take it for reading and it doesn't modify this list. destroy_context() reaps this list lockless, this is OK. This patch assumes that sram_free() doesn't need to be serialized (afaics it does the locking correctly), and it is safe to call sram_free(addr) after its sram_list_struct was removed from list. Signed-off-by: Oleg Nesterov <oleg(a)redhat.com> --- arch/blackfin/mm/sram-alloc.c | 24 ++++++++++++++++-------- 1 file changed, 16 insertions(+), 8 deletions(-) --- 34-rc1/arch/blackfin/mm/sram-alloc.c~IUAV_2_SRAM_LIST_LOCKING 2010-05-27 17:51:48.000000000 +0200 +++ 34-rc1/arch/blackfin/mm/sram-alloc.c 2010-05-27 21:31:01.000000000 +0200 @@ -702,19 +702,23 @@ EXPORT_SYMBOL(l2_sram_free); int sram_free_with_lsl(const void *addr) { - struct sram_list_struct *lsl, **tmp; + struct sram_list_struct *lsl = NULL, **tmp; struct mm_struct *mm = current->mm; + down_write(&mm->mmap_sem); for (tmp = &mm->context.sram_list; *tmp; tmp = &(*tmp)->next) - if ((*tmp)->addr == addr) - goto found; - return -1; -found: - lsl = *tmp; + if ((*tmp)->addr == addr) { + lsl = *tmp; + *tmp = lsl->next; + break; + } + up_write(&mm->mmap_sem); + + if (!lsl) + return -1; + sram_free(addr); - *tmp = lsl->next; kfree(lsl); - return 0; } EXPORT_SYMBOL(sram_free_with_lsl); @@ -749,10 +753,14 @@ void *sram_alloc_with_lsl(size_t size, u kfree(lsl); return NULL; } + + down_write(&mm->mmap_sem); lsl->addr = addr; lsl->length = size; lsl->next = mm->context.sram_list; mm->context.sram_list = lsl; + up_write(&mm->mmap_sem); + return addr; } EXPORT_SYMBOL(sram_alloc_with_lsl); -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo(a)vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/ |