qemu-stable
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH for-9.1] linux-user: Handle short reads in mmap_h_gt_g


From: Philippe Mathieu-Daudé
Subject: Re: [PATCH for-9.1] linux-user: Handle short reads in mmap_h_gt_g
Date: Fri, 16 Aug 2024 07:57:45 +0200
User-agent: Mozilla Thunderbird

On 15/8/24 23:32, Richard Henderson wrote:
In particular, if an image has a large bss, we can hit
EOF before reading all host_len bytes of the mapping.

Cc: qemu-stable@nongnu.org
Fixes: eb5027ac618 ("linux-user: Split out mmap_h_gt_g")
Resolves: https://gitlab.com/qemu-project/qemu/-/issues/2504
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
  linux-user/mmap.c | 19 +++++++++++++++----
  1 file changed, 15 insertions(+), 4 deletions(-)

diff --git a/linux-user/mmap.c b/linux-user/mmap.c
index 6418e811f6..de9ab13754 100644
--- a/linux-user/mmap.c
+++ b/linux-user/mmap.c
@@ -853,10 +853,21 @@ static abi_long mmap_h_gt_g(abi_ulong start, abi_ulong 
len,
      }
if (misaligned_offset) {
-        /* TODO: The read could be short. */

I note there are other short reads in {linux,bsd}-user/mmap.c.

Reviewed-by: Philippe Mathieu-Daudé <philmd@linaro.org>

-        if (pread(fd, p, host_len, offset + real_start - start) != host_len) {
-            do_munmap(p, host_len);
-            return -1;
+        size_t o = 0;
+        while (1) {
+            ssize_t r = pread(fd, p + o, host_len - o,
+                              o + offset + real_start - start);
+            if (likely(r == host_len - o) || r == 0) {
+                /* Complete or EOF */
+                break;
+            }
+            if (unlikely(r == -1)) {
+                /* Error */
+                do_munmap(p, host_len);
+                return -1;
+            }
+            /* Short read -- iterate */
+            o += r;
          }
          if (!(host_prot & PROT_WRITE)) {
              mprotect(p, host_len, host_prot);




reply via email to

[Prev in Thread] Current Thread [Next in Thread]