As subject suggests generic scsi module from 2.4.19 corrupts data.
Backing back to 2.4.18 code [as suggested below] fixes the problem.
as the code appears equivalent. Corruption occurs if you break out
the inner loop few lines later:
		if (ksglen > usglen) {
		    ...
		    p += usglen;
		    ksglen -= usglen;
		    break;
		}
and reenter it. Then ksglen and p are reset which results in some
duplicate data injection.
Cheers. Andy.
--- ./drivers/scsi/sg.c.orig	Sat Aug  3 02:39:44 2002
+++ ./drivers/scsi/sg.c	Wed Aug 21 00:44:16 2002
@@ -1884,11 +1884,16 @@
 	    res = sg_u_iovec(hp, iovec_count, j, 1, &usglen, &up);
 	    if (res) return res;
 
+#if 0
 	    for (; k < schp->k_use_sg; ++k, ++sclp) {
 		ksglen = (int)sclp->length;
 		p = sclp->address;
 		if (NULL == p)
 		    break;
+#else
+	    for (; (k < schp->k_use_sg) && p;
+		++k, ++sclp, ksglen = (int)sclp->length, p = sclp->address) {
+#endif
 		ok = (SG_USER_MEM != mem_src_arr[k]);
 		if (usglen <= 0)
 		    break;
@@ -2040,11 +2045,16 @@
 	    res = sg_u_iovec(hp, iovec_count, j, 0, &usglen, &up);
 	    if (res) return res;
 
+#if 0
 	    for (; k < schp->k_use_sg; ++k, ++sclp) {
 		ksglen = (int)sclp->length;
 		p = sclp->address;
 		if (NULL == p)
 		    break;
+#else
+	    for (; (k < schp->k_use_sg) && p;
+		++k, ++sclp, ksglen = (int)sclp->length, p = sclp->address) {
+#endif
 		ok = (SG_USER_MEM != mem_src_arr[k]);
 		if (usglen <= 0)
 		    break;
-
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/